# VideoDraft docs, every page # VideoDraft documentation > VideoDraft ADE is a Mac app where AI agents help you make videos and build software. You can also use the same media tools in your browser, in other AI apps and in a terminal. Source: https://videodraft.ai/docs/overview ## What you can do | You want to | VideoDraft can | Start with | | --- | --- | --- | | Create media | Images, video, voice, music, sound effects, dialogue, dubbing, presenter videos and 3D models, from text or your own references. | [AI Studio](https://videodraft.ai/docs/web-ai-studio) or [VideoDraft mode](https://videodraft.ai/docs/videodraft-mode) | | Edit a video | A real timeline on your Mac: cuts, speed ramps, transitions, captions, color, sound, multicam, and export to MP4, ProRes, Premiere Pro, DaVinci Resolve or Final Cut Pro. | [VideoDraft Editor](https://videodraft.ai/docs/native-editor) | | Plan a production | Scripts, storyboards, animatics and full AI productions in your browser. | [Scripts and storyboards](https://videodraft.ai/docs/web-storyboards) | | Direct AI agents | Claude Code, Codex, Grok Build, OpenCode or VD Code. Run tasks side by side, plan before building, schedule repeat jobs, let an agent use your apps and continue from your phone. | [Work with agents](https://videodraft.ai/docs/working-with-agents) | | Build software | A branch per task, diffs, AI code review and pull requests. | [Code mode](https://videodraft.ai/docs/code-mode) | | Use another AI app | Make media from Claude, ChatGPT, Codex, Cursor or a terminal script. | [MCP](https://videodraft.ai/docs/mcp) or [CLI](https://videodraft.ai/docs/cli) | ## Where your work lives | Item | What it is | Where | | --- | --- | --- | | ADE project and task | A project groups tasks. A task is one piece of work with its own chats and files. | Your Mac | | Editor project (.vdproject) | A VideoDraft Editor timeline and its media list. | Your Mac | | AI Studio project | A collection of images, video and sound you made. A few menus call it a session. | VideoDraft cloud | | Video project | A script, storyboard and timeline in the web app. | VideoDraft cloud | Nothing syncs between them on its own. To move media, download it from one and import it into the other. ## Learn by doing Each tutorial ends with something you can watch, run or share. - [Your first video](https://videodraft.ai/docs/first-video) - [Your first code change](https://videodraft.ai/docs/first-code-change) - [Your first image and clip](https://videodraft.ai/docs/web-quickstart) - [All tutorials](https://videodraft.ai/docs/tutorials) --- # All tutorials > Step-by-step guides that end with something you can watch, run or share. Tutorials for the guide you are in come first. Source: https://videodraft.ai/docs/tutorials ## On your Mac: VideoDraft mode | Tutorial | You finish with | Cost | | --- | --- | --- | | [Your first video](https://videodraft.ai/docs/first-video) | A 15-second video made from your own images or clips | Agent usage only | | [A narrated video, start to finish](https://videodraft.ai/docs/editor-walkthrough) | Your voiceover and clips, edited with captions and music | Agent usage only | | [Keep a character or product consistent](https://videodraft.ai/docs/consistent-characters) | Three shots that match | Media credits | | [Bring web media into ADE](https://videodraft.ai/docs/web-to-ade) | A clip from the web app placed in your edit | Agent usage only | | [Scheduled tasks](https://videodraft.ai/docs/scheduled-tasks) | A tested prompt that runs on a schedule | Agent usage per run | ## On your Mac: Code mode | Tutorial | You finish with | Cost | | --- | --- | --- | | [Your first code change](https://videodraft.ai/docs/first-code-change) | A reviewed change on its own branch | Agent usage only | | [Computer Use](https://videodraft.ai/docs/computer-use#first-task) | An agent working in your browser while you watch | Agent usage only | ## In your browser | Tutorial | You finish with | Cost | | --- | --- | --- | | [Your first image and clip](https://videodraft.ai/docs/web-quickstart) | A product photo and a short clip, downloaded | Media credits | | [Produce and export a video](https://videodraft.ai/docs/web-production) | A finished video from your storyboard | Media credits, Pro plan for AI Production | | [Make an avatar video](https://videodraft.ai/docs/web-avatar-video) | A presenter video with captions | Media credits | | [Speech, music and sound](https://videodraft.ai/docs/web-sound) | Narration, dialogue, music and your own cloned voice | Media credits | ## From another AI app or a terminal | Tutorial | You finish with | Cost | | --- | --- | --- | | [MCP connector](https://videodraft.ai/docs/mcp#first-result) | An image made from Claude, ChatGPT, Codex or Cursor | Media credits | | [Command line (CLI)](https://videodraft.ai/docs/cli) | Results downloaded by a script | Media credits | --- # Install ADE > Install VideoDraft ADE on your Mac, connect an agent and pick your first tutorial. Source: https://videodraft.ai/docs/getting-started ## What you need - An Apple silicon Mac with macOS 15 or later. VideoDraft Editor comes with ADE. - A VideoDraft account. - An agent. VD Code comes with your VideoDraft plan. You can also connect Claude Code or Codex with your own sign-in, Grok Build, or OpenCode with your own API key. ## Install and connect - [Download for Mac](https://videodraft.ai/download/mac) 1. Download VideoDraft and move it to Applications. 2. Open VideoDraft and sign in to your VideoDraft account. 3. Open Settings > Providers and connect an agent, or choose VD Code. ADE installs what the agent needs. > **Claude Code and Codex use your existing sign-in** > > If you already use them in Terminal, ADE picks up that sign-in. Otherwise run claude auth login or codex login, or sign in from Settings > Providers. ## Pick your first tutorial Pick a mode before you start a chat: click the mode name at the top of the sidebar, or press Cmd + O. A chat keeps the mode it started in. - [Your first video](https://videodraft.ai/docs/first-video) - [Your first code change](https://videodraft.ai/docs/first-code-change) ## Keep VideoDraft up to date VideoDraft checks for updates on its own, downloads them in the background and asks to restart when one is ready. To check now, choose VideoDraft > Check for Updates… in the menu bar, or click Check now in Settings > Updates. --- # VideoDraft mode > Plan a production, make the media and finish the edit on your Mac. The website calls this Vibe Film. Source: https://videodraft.ai/docs/videodraft-mode ## Start a video project Bring an idea, a script, reference images or footage. Agents plan the video and make the media with you. Then you both edit the same timeline in VideoDraft Editor. - [Make your first video](https://videodraft.ai/docs/first-video) - [Install ADE](https://videodraft.ai/docs/getting-started) ## Find your way around ![VideoDraft mode with a production plan, a character reference and project files side by side.](https://videodraft.ai/ade-screenshots/home-workspace.webp) A production in VideoDraft mode, with the plan, a character reference and the task’s files open at once. | Area | Use it for | | --- | --- | | Left sidebar | New task, Dashboard, Scheduled and your projects and tasks. The mode name at the top switches modes. | | Center | Chats, plans and file previews. Split it to see two at once. | | All files | Open your brief, look at images and play clips or audio. | | Activity | See what the agent is doing. | | Editor | Create or link the task’s VideoDraft Editor project, then click Open in Editor. | ## What you can make ![VideoDraft mode with a generated clip in the chat, a generated still in the center and a sound effect preview on the right.](https://videodraft.ai/ade-screenshots/home-generate.webp) Media the agent makes lands in the task’s media folder. Open any file from All files to watch or listen. | You want | Ask the agent for | Where it ends up | | --- | --- | --- | | New media | Images, video, voice, music, sound effects, dubbing or upscales. | Files in media/ | | An edit | A timeline built from your media: cuts, captions, color, sound and export. | A .vdproject in [VideoDraft Editor](https://videodraft.ai/docs/native-editor) | | A web video project | A script, storyboard and production to continue in the browser. Ask for it by name. | A [video project](https://videodraft.ai/docs/web-production) in the web app | Example: ``` Make a 6-second 16:9 establishing shot from @media/reference.png. Keep the camera still, add slow drifting mist and keep the product’s shape. Estimate the cost first, save the clip in media/ and show it here. ``` > **Check models before a big batch** > > Models change often. Ask the agent which ones are available and what the job costs before it makes many shots. ## From brief to finished video 1. Write a brief: audience, length, aspect ratio and the result you want. 2. Agree on a look with a few reference images before making every shot. 3. Make a small first batch and look at the actual files. 4. Put the chosen media in VideoDraft Editor and adjust timing, captions, sound and color. 5. Watch the whole cut, then export it or hand it to another editor. - [Keep characters consistent](https://videodraft.ai/docs/consistent-characters) - [Work with agents](https://videodraft.ai/docs/working-with-agents) --- # Code mode > Build software with agents, with files, terminals, diffs and Git beside the chat. The website calls this Vibe Code. Source: https://videodraft.ai/docs/code-mode ## Start with a repository Open a folder or repository, or clone one from GitHub. Give the agent one concrete change, then run it and review the diff. - [Make your first code change](https://videodraft.ai/docs/first-code-change) - [Install ADE](https://videodraft.ai/docs/getting-started?mode=code) ## Keep the result beside the chat ![Code mode showing a source diff, the changed files list and a running development server.](https://videodraft.ai/ade-screenshots/code-review.webp) Code mode keeps the diff and the Run output beside the chat. | Tab or control | Use it for | | --- | --- | | All files | Open, preview and edit files. | | Changes | See what changed compared with the branch the task started from. | | Activity | See what the agent is doing. | | Git | Branch, checks and pull request controls. | | Setup, Run, Terminal | Setup output, your run command and shells in the task folder. | | Review, Create PR | Buttons in the right panel header for an AI review and a pull request. | ## Give each task its own branch When you open a Git repository, choose Isolated workspaces. Each task then gets its own branch and copy of the files, so several agents can work at once. - [How tasks share files](https://videodraft.ai/docs/projects-and-workspaces?mode=code) ## Build, check and review 1. Ask for a plan first when the approach needs agreement. 2. Make one change and run the project’s checks. 3. Try the result and read the diff. Comment on lines that need work. 4. Click Review for an AI review and fix what it finds. 5. Commit, push and open a pull request. - [Git and pull requests](https://videodraft.ai/docs/git-and-pull-requests) - [Work with agents](https://videodraft.ai/docs/working-with-agents?mode=code) --- # Your first video > Turn three of your own images or clips into a 15-second video with an agent, change it, and export an MP4. Source: https://videodraft.ai/docs/first-video ## What you need Three images, or clips at least 5 seconds long, that you already have. Product photos, screenshots and phone videos all work. Your own files cost no media credits. The agent’s work still counts toward your provider plan or VD Code weekly limit. 1. Make a folder for the project, for example first-video, with a folder called media inside it. 2. Copy your three files into media/. Name them shot-1, shot-2 and shot-3 and keep their extensions, so the order is clear. ## Open the folder in ADE 1. In ADE, switch to VideoDraft mode: click the mode name at the top of the sidebar. 2. Click the folder icon next to Projects in the sidebar and choose Open folder…. Pick your first-video folder. 3. Start a new task in the project and open its chat. 4. Open All files and check the three files are in media/. ## Ask for the first cut Paste into the chat: ``` In VideoDraft Editor, build a 15-second, 16:9, 1080p, 30 fps video from media/shot-1, media/shot-2 and media/shot-3, in that order. Give each one 5 seconds, show each image or clip whole without cropping, and use straight cuts. Mute any sound from the clips. Save the editor project in this folder and show me the editor when it is ready. ``` The prompt sets the length, size, order and scope, so the agent has nothing to guess. If its plan adds extra media or a web project, ask it to stick to the brief. ## Check the timeline ![VideoDraft Editor with the Library, preview, Inspector and timeline.](https://videodraft.ai/ade-screenshots/film-editor.webp) VideoDraft Editor. Your three shots sit on the timeline along the bottom, and the preview in the middle plays them. 1. In the right panel, open the Editor tab. It shows the linked project. Click Open in Editor. 2. Check the project is 1920 × 1080 at 30 fps. 3. Check the shots start at 0, 5 and 10 seconds and end at 15. 4. Press Space to play it. Each shot should be whole in the frame, with no gaps and no sound. > **If the editor window does not appear** > > Ask the agent to “Show VideoDraft Editor”. An agent can edit with the window closed, so the window only opens when asked. Then ask for a change: ``` Change the timing to 4 seconds for shot 1, 6 seconds for shot 2 and 5 seconds for shot 3. Keep the total at 15 seconds. Show me the timeline before exporting. ``` ## Export and watch the file Finish: ``` Export this timeline as H.264 1080p to media/first-video.mp4. Wait until the export finishes, check the file and give me a link to it. ``` Open first-video.mp4 from All files or the link in the chat and watch it to the end. The job is done when the file plays, not when the export starts. ## Part 2: add a shot made with AI Now swap the first shot for one made with AI. This part uses media credits, so the agent estimates the cost before making anything. Paste into the chat: ``` Propose one visual for the first 4 seconds of this video. Check the available image and video models and estimate the cost. After I approve, make a still image first and show it to me. Once I pick it, animate it, save it in media/ and put it in the first slot without changing the rest of the timeline. ``` - [A narrated video, start to finish](https://videodraft.ai/docs/editor-walkthrough) - [How credits work](https://videodraft.ai/docs/providers-and-usage) --- # Your first code change > Open a sample website on its own branch, improve it with an agent and review the diff. Source: https://videodraft.ai/docs/first-code-change ## Open the sample [Download the sample site](https://videodraft.ai/docs-assets/sample-site.zip) A one-page website in a small Git repository, made for this tutorial. Free, no account needed. 1. Unzip the download. You get a sample-site folder: a small Git repository with one commit. 2. In ADE, switch to Code mode: click the mode name at the top of the sidebar. 3. Click the folder icon next to Projects in the sidebar and choose Open folder…. Pick sample-site. 4. When ADE asks how to open the repository, choose Isolated workspaces. Each task then gets its own branch. 5. Start a new task in the project and open its chat. ## Run it First prompt: ``` Read index.html and explain what this page does. Start a local static server on a free port, give me the address and keep the server running. Do not change any files yet. ``` Open the address in your browser. You should see the Northwind sample page with three sections. To start the server with the Run button next time, save a run command in [Project settings](https://videodraft.ai/docs/project-settings). ## Make one change and review it Second prompt: ``` Make keyboard focus easy to see on this page. Keep the content and layout. Explain the change, edit the CSS, then check that pressing Tab clearly highlights each link. ``` 1. Reload the page and press Tab to move through the links. 2. Open the Changes tab in the right panel and read the CSS the agent changed. 3. Click Review in the right panel header. A new chat reviews the change and adds comments to the diff. 4. Ask the first chat to fix anything the review found. > **Only one chat should edit** > > Chats in the same task share its files. Let the reviewer read and comment while the first chat makes the edits. ## Keep the change You are done when the page still works, focus is easy to see and you understand the CSS change. Ask the agent to stop the server. To keep it, ask the agent to commit the change on this task’s branch. Commit & Push and Create PR need a GitHub remote. To try a pull request, push the sample to your own GitHub repository first. - [Git and pull requests](https://videodraft.ai/docs/git-and-pull-requests) - [Work with agents](https://videodraft.ai/docs/working-with-agents?mode=code) --- # Keep a character or product consistent > Pick one reference image, use it for three shots and check they match before you animate. Source: https://videodraft.ai/docs/consistent-characters ## Before you start > **This tutorial uses media credits** > > Every image and clip costs credits. The agent estimates before it makes anything. You need one clear reference image of your character or product that you have the right to use. Write down what must never change: face, clothing, packaging, colors, small details. Save the image as media/references/main.png. No reference yet? Make one first: ``` Make one reference image of an original character: a woman in her thirties with short black hair, a blue wool coat and a red scarf, full body, facing the camera, plain light gray background. Estimate first. Save it as media/references/main.png. ``` ## Make three matching stills Paste into the chat: ``` Using media/references/main.png, make three stills of the same character: walking into a café, sitting by a window, and a close-up smiling. Keep the face, red scarf and blue coat exactly the same. Change only the camera angle and action. First show me the shot list and the cost. ``` 1. Approve one still first and check the face, hands, clothing and background. 2. Fix the reference or the prompt before making the other two. 3. Save the stills you like in media/references/ with a short note of what must match. ## Animate the stills Next prompt: ``` Animate the three stills I picked, 4 seconds each. Keep the look and composition; describe only the action and camera move. Check which input types and lengths the chosen model supports. Estimate before you start and save the clips in media/. ``` Models take images in different ways: as the first frame, as first and last frames, or as a reference for the look. References help a lot, but check every clip. ## Compare the shots - Put the clips side by side and compare face or product details, clothing, size and direction of movement. - If a clip has sound, listen for changes in voice or room tone. - Remake a shot that drifts from the reference, then compare again. - Put only the shots that match in your edit. In the web app, [Character Studio](https://videodraft.ai/docs/web-character-studio) and [Product Studio](https://videodraft.ai/docs/web-product-studio) save a character or product so you can reuse it in any shot. --- # Bring web media into ADE > Find a clip you made in the web app, download it into a task and place it in your edit. Source: https://videodraft.ai/docs/web-to-ade ## Find and download the clip Nothing syncs between the web app and your Mac on its own. You download the clip, then import it into your editor project. 1. Open AI Studio and pick the project where you made the clip from the Projects list. If an agent connected over MCP made it, ask that agent which AI Studio project it used. 2. Click the clip to preview it and make sure it is the one you want. 3. Hover the clip and click Download. Save it in the task’s media/ folder. 4. To remake the clip later, choose Copy Prompt from its menu and save the prompt. ## Put it in your edit Paste into the ADE chat: ``` Import media/selected-shot.mp4 into the linked VideoDraft Editor project. Check its size, length and audio. Place it in the opening slot without moving the rest of the edit, and show me the result. ``` Check the Editor tab shows the right project, then watch the timeline and the export. If you move the file after importing it, move it back or relink it before exporting. - [VideoDraft Editor](https://videodraft.ai/docs/native-editor) - [AI Studio basics](https://videodraft.ai/docs/web-ai-studio) --- # Work with agents > Give clear tasks, pick how the agent works, correct it while it runs and hand work between chats. Source: https://videodraft.ai/docs/working-with-agents ## Give a clear task Say what you want, what to use, the limits, where to save it and how you will judge it. Keep decisions that should last in a file, such as brief.md, so any later chat can read them. A task that works for most jobs: ``` Outcome: [what to make] Inputs: [files or linked folders] Limits: [length, format, style, budget] Save to: [folder or file] Done when: [what you will check] Ask me before: [decisions you want to make] ``` ## Choose how the agent works The message box has a model picker, a thinking level and a mode button that reads Auto, Plan or Ask. | Control | Use it when | | --- | --- | | Model | Pick the agent and model for the job. See [which agent suits what](https://videodraft.ai/docs/providers-and-usage#connect). | | Thinking | Raise it for hard planning or debugging. | | Plan | You want to see the approach first. The agent changes nothing until you choose Approve or Hand off. | | Ask | You want the agent to ask before protected actions, such as running a command. | | Auto | The task is clear and the agent should work without stopping for approval. | > **Two different kinds of mode** > > VideoDraft mode and Code mode set up the whole app for video or software. Auto, Plan and Ask set how one chat’s agent works. You can plan a video or run a code task on Auto. ## Steer, queue or stop | Action | Example | | --- | --- | | Steer | “Keep the red scarf.” Your message goes into the work that is running now. | | Queue | “After this, make a vertical version.” Your message waits until the current turn ends. | | Stop | End the current work, then look at what it finished before you continue. | Choose which one happens by default in Settings > General > Follow-up while streaming. The message box shows the shortcut for the other one, such as Cmd + Enter to steer. Slash commands always wait in the queue. ## Plan first, then hand off 1. Set the mode button to Plan and describe the task. 2. When the bar says Plan ready, open the plan and read it. 3. Choose Approve to build it in this chat, or Hand off to start a new chat that builds it. A new chat shares the task’s files but starts a fresh conversation. 4. In any new, empty chat, the Add plans and Add chat transcripts chips above the message box attach earlier plans or chats as files. ## Run work side by side ![The ADE Dashboard progress board, with the running agents list open at the top of the window.](https://videodraft.ai/ade-screenshots/code-parallel.webp) Click the agent count at the top of the window to see every chat that is working. Dashboard > Progress board shows where each task stands. 1. Keep one deliverable and its revisions in one task. 2. Start another task for separate work. For a video, one task can explore shots while another prepares sound. 3. Start a new chat for a reviewer or a different agent. Say whether it should only look or also edit. 4. Link only the folders a task needs. > **Shared folders are not locked** > > A task opened with Isolated workspaces has its own copy of the files. Everything else is shared: if two agents edit the same file or .vdproject, one can overwrite the other. Say which chat edits what. ## Keep long chats on track The small ring in the message box shows how much more the chat can hold. Hover it for details. This is separate from usage limits and credits. Type /compact to have the agent summarize the chat and make room. Before you move to a new chat, ask the agent for a short handoff note: the goal, the decisions made, the files, what is done and the next step. To reuse a decision from another chat, just ask. Agents can search and read your other chats. Example: ``` Find the chat where we picked the character design and use that reference here. Tell me if any later chat changed it. ``` ## Go back or try again A turn is one message and the agent’s full reply. Hover a turn to see its actions. | Action | What it does | | --- | --- | | Copy | Copies the turn. | | Fork | Fork to new tab or Fork to new task starts a new chat from this point, with the earlier chat attached. | | Reset to this point | Rolls the task’s files back to this turn and removes later turns. See [what Reset can undo](https://videodraft.ai/docs/troubleshooting#reset). | | Retry | On a failed turn, runs it again. | | Continue | On a turn you stopped, picks up where it left off. | ## Check the result - Look at the real file or running page, not only the agent’s summary. - For code, read the diff and the checks. For video, watch the whole exported file. - Write down the chosen direction and what is left. - Start a fresh chat when the goal changes, and bring only what it needs. - [Save reusable prompts](https://videodraft.ai/docs/prompt-library) - [Fix a problem](https://videodraft.ai/docs/troubleshooting) --- # Projects and tasks > Organize work so every agent has the right files and a clear scope. Source: https://videodraft.ai/docs/projects-and-workspaces ## Projects hold related work A project is a top-level item in the sidebar. Use one for a film, campaign, client or code repository. Click the folder icon next to Projects to add one. ![ADE with the Projects menu open, showing New project, Open folder and Clone GitHub repo.](https://videodraft.ai/ade-screenshots/code-context.webp) The Projects menu. New project makes a project ADE manages, Open folder… uses a folder you already have, and Clone GitHub repo… copies a repository to your Mac. | When you open | What happens | | --- | --- | | A folder without Git | ADE works in the folder itself. The Changes and Git tabs stay empty. | | A Git repository with Isolated workspaces | Each task gets its own branch and copy of the files, so agents can work at the same time. The repository needs at least one commit. Code mode suggests this. | | A Git repository with Edit this folder directly | Every task works in the folder itself. VideoDraft mode suggests this. | ## Give each outcome its own task A task is one piece of work with its own chats and files. Some menus call a task a workspace, such as Link workspaces. Keep revisions of one deliverable in one task, and start a new task when the goal changes. | Keep in one task | Start another task for | | --- | --- | | A teaser and its revisions | A different client’s campaign | | One storyboard and its shots | An unrelated music video | | A weekly social post series | A one-off product launch | ## Where files go - Put scripts, briefs, shot lists and notes in the task folder. - Put images, video and audio in media/. Example task folder: ``` launch-teaser/ ├── brief.md ├── treatment.md ├── shot-list.csv └── media/ ├── keyframe-01.png ├── teaser-v1.mp4 └── voiceover.wav ``` ## Link more folders An agent can read and write outside its task once you link a folder. Click + in the message box and choose Link workspaces, or type /add-dir. Pick another task from any project, or any folder on your Mac. - Linked folders show as chips above the message box. Remove one there. - Links belong to the task, so every chat in it can use them, even after a restart. ## Choose which mode lists a project To list a project or task in VideoDraft mode, Code mode or both, right-click it and choose Show in. A task can also use its project’s setting. --- # Prompts, instructions and skills > Save prompts you send often, and teach agents how your project works. Source: https://videodraft.ai/docs/prompt-library ## Pick the right place | Use | For | | --- | --- | | Prompt Library | A prompt you want to reuse, edit and send again. | | Project instruction files | Rules every agent should follow in a project. Each agent reads its own file, such as AGENTS.md or CLAUDE.md. | | Skill | A step-by-step procedure an agent loads when a task needs it. | | Project settings | Setup and run commands for code projects. | ## Save and reuse a prompt 1. Press Cmd + P, or click Prompts in the message box. 2. Click New prompt. Give it a title and the text, then click Add prompt. 3. Click a saved prompt to put it in the message box. Long prompts attach as a text file. 4. Edit the details, then send. > **Choosing a prompt does not send it** > > It only fills the message box, so you can check it first. ## A prompt worth saving Review a cut: ``` Review the exported video at [path]. Check the opening, pacing, continuity, readable text, audio and last frame. List the few changes that would help most. Do not change the timeline until I pick the changes. ``` Mark the parts you replace each time with brackets. Never put passwords or keys in saved prompts or instruction files. ## Skills A skill is a folder of instructions an agent loads when a task needs it. ADE installs two when it starts, and every agent can use them. | Skill | What it teaches | | --- | --- | | videodraft | Making video, images, audio and 3D, and working in VideoDraft Editor. | | videodraft-ade | Reading other chats and tasks in ADE. | To add your own, make a folder with a SKILL.md file inside a skills folder, such as .claude/skills in your project. It shows up when you type /. --- # Accounts, usage and credits > Connect an agent, see what pays for what, and check how much is left. Source: https://videodraft.ai/docs/providers-and-usage ## What pays for what | Work | Paid from | | --- | --- | | Claude Code, Codex or Grok Build chats | Your account with that provider. | | OpenCode chats | The API key you connected. | | VD Code chats | The VD Code weekly limit in your VideoDraft plan. | | Images, video and sound made by VideoDraft | Your VideoDraft credits, whichever agent or app asked for them. | | Media made with your own ElevenLabs or Fal key | That provider account, for the models that support it. | | Editing and exporting in VideoDraft Editor | Nothing. It runs on your Mac. | ## Connect an agent in ADE Open Settings > Providers. Each agent has its own tab, and you can connect several. Pick models inside a chat. | Agent | How to connect | Good for | | --- | --- | --- | | Claude Code | Your Claude sign-in (claude auth login) or an API key. | Long creative briefs, making lots of media, Plan mode and remote control. | | Codex | Your Codex sign-in (codex login) or an API key. | Careful step-by-step work, long-running goals, second opinions and Computer Use. | | Grok Build | Sign in with Grok. | Grok models, with your Grok usage shown in ADE. | | OpenCode | An API key from Google AI Studio, DeepSeek, xAI, OpenRouter or OpenCode Zen. | Models you already pay for. | | VD Code | Your VideoDraft account. | Included with your plan. No other subscription needed. | A chat stays with the agent it started with. Picking another agent’s model opens a new chat in the same task. > **Enterprise data privacy turns some off** > > When your company manages Enterprise data privacy for your VideoDraft account, VD Code and Grok Build are unavailable. ## VD Code weekly limit - VD Code has a weekly limit, separate from your credits. - The 7-day window starts with your first use. When it ends, the full limit comes back. - Higher plans have higher limits. The pricing page lists them. - At the limit, the chat stops with Usage limit hit. Wait for the window to reset, upgrade, or continue in a new chat with another agent. - [Plans and limits](https://videodraft.ai/pricing) ## See what is left | Where | What it shows | | --- | --- | | Usage percentage, bottom of the ADE sidebar | How much of your current limit is used. Click it for Usage Limits. | | Usage Limits | A tab per connected agent. The VideoDraft tab shows your credits and VD Code weekly limit, with Buy credits. | | Settings > Account | Your plan, credits and VD Code limit, with Manage plan and Buy credits. | | Dashboard | Daily activity, token usage over time and usage by model. | | In AI Studio: Project Settings > Usage | What one AI Studio project has spent. | ## Check the cost before a batch - In AI Studio, the Create button shows the cost for your current settings, such as “12 cr”. - Cost depends on the model, length, resolution, audio and references. - Agents and the CLI can estimate a job first. Ask for one before anything large. - Make one shot, check it, then make the rest. - A sign-in error or a provider limit is not a credits problem. Fix it in Settings > Providers. ## Use your own ElevenLabs or Fal key Add keys in the web app under Settings > API Keys. This is separate from connecting an agent in ADE. - With a Fal key on, supported models bill your Fal account, and models it cannot run are turned off. - Voices you cloned in ElevenLabs need the ElevenLabs key that owns them. - Check the usage record to confirm which account paid. --- # Scheduled tasks > Test a prompt, schedule it, and review every run. Source: https://videodraft.ai/docs/scheduled-tasks ## Try the prompt first A scheduled run cannot ask you anything. Run the prompt in a normal chat until the result is right. Name the inputs, where to save, what done looks like and what to do when an input is missing. Video example: ``` Read the brief and the new files in the linked intake folder. Write three social video ideas to a dated Markdown file, with the source file names. Flag missing references. Do not make paid media. If there is nothing new, save a short note saying so. ``` Code example: ``` Read the project instructions and the repository. Write a dated summary of recent changes and any checks that need attention. Do not edit files, commit, push or open a pull request. ``` ## Create the schedule 1. Open Scheduled in the sidebar and click New task. 2. Choose the Mode, then fill in Name and Prompt. Pick the agent and model. 3. Choose a Schedule: Once, Hourly, Daily, Weekdays or Weekly. Set Run at and check the time zone. 4. Choose the Project and which Task each run uses (see the table). 5. Click Create. Then open the scheduled task’s menu and choose Run now to test a real run. 6. Open the run’s chat and output file. Fix the prompt if anything is missing. ## Choose where each run works | Task option | Use it for | | --- | --- | | New task each run | Separate results, each in a fresh task folder. | | Create new, then reuse | Repeat work that should keep its files between runs. | | Use existing task | Work that belongs in a task you already have. A run can start while you are working in that task. | A Once schedule offers Create new task or Use existing task. > **Fresh tasks start empty** > > A file saved in an earlier task is not in a new one. Link the input folder, and check the first run can read it. ## How runs behave - Runs use Auto mode, so they never stop for approval. - Each run makes a chat you can open from Scheduled. Turn on Show runs in sidebar to list them there too. - ADE must be open for a run to start on time. Turn on Keep computer awake at the top of Scheduled, or in Settings > General. The display can still sleep. - If ADE was closed at run time, it runs the task once when it opens again, even if several run times were missed. - A Codex run can use Computer Use only when Allow Computer Use with Codex Auto is on and the Mac is unlocked. ## Good jobs to schedule - Three social video ideas every Monday from a standing brief. - A daily summary of new footage in an intake folder. - A weekly list of shots still missing references. - A one-time overnight batch job. --- # Computer Use > Let a Codex chat use Mac apps and your browser tabs while you watch. Source: https://videodraft.ai/docs/computer-use ## What it is Computer Use lets a Codex chat open and use apps on your Mac, including tabs in your own browser. It clicks, types and reads the screen, and you watch it in a live view in the chat. - Codex chats only. - Off until you turn it on. - Needs macOS 15 or later and an unlocked Mac. ## Turn it on 1. Open Settings > Computer use and turn on Enable Computer use. 2. Under Mac permissions, choose Grant access for Screen Recording, Accessibility and Background input. macOS lists the helper as VideoDraft Computer Use. 3. Choose Check permissions to confirm all three. 4. Pick App access. Auto lets Codex use apps without asking. Ask checks with you first. - Keep apps in the background is on by default, so Codex can work without bringing apps to the front. - With App access set to Ask, Always allowed apps lists the apps you approved for good. ## Use your own browser The VideoDraft Browser Companion extension lets Codex work in your browser tabs in the background. It works in Chrome, Edge, Brave, Opera and Vivaldi. 1. In Settings > Computer use, under External browsers, choose your Browser, then Set up next to Background browser access. Finder opens the extension folder. 2. In the browser, open the Extensions page, turn on Developer mode, choose Load unpacked and select that folder. 3. In the companion, choose Allow this profile to connect, then Copy token. 4. Paste the token into Connection token in ADE and choose Save. ADE keeps it in your Keychain. 5. Send a new message in the chat so it picks up the browser tools. ## Try a first task First browser task: ``` Use my connected browser to open videodraft.ai/docs. Find the Install ADE page and list what I need before installing. Do not sign in, submit forms or change settings. ``` 1. Watch the live view and check Codex uses the browser you expect. 2. If the tools are missing, ask the agent to check Computer Use setup. It reports what is missing. ## Controls while it works | Control | What it does | | --- | --- | | Stop task | Stops the work right away. | | Hide app view | Hides the live view without stopping the work. Show app view brings it back. | | Move to floating window | Moves the live view into its own window. | | Show in chat | Puts the live view back in the chat. | | Hide for this task | Keeps the live view hidden for the rest of the task. | With App access set to Ask, Codex asks before using an app. Choose Allow for this task, Always allow or Decline. ## What it will not do - It cannot control ADE, the Codex app, terminal apps such as Terminal, iTerm2, Ghostty and Warp, or System Settings. - It does not work while the Mac is locked, and says so in the chat. - It is not available through remote control. - Plan mode and Enterprise data privacy turn it off. - Scheduled runs can use it only with Allow Computer Use with Codex Auto turned on, and cannot approve new apps. > **Not the same as browser preview** > > The in-app browser preview shows local pages inside ADE. Computer Use operates other apps and your own browser. They have separate settings and permissions. --- # Remote control > Continue a Claude chat from your phone or browser while it keeps running on your Mac. Source: https://videodraft.ai/docs/remote-control ## How it works Remote control lets you carry on a Claude chat from your phone or a browser. The chat does not move: it keeps running on your Mac with your files and tools, and your phone becomes another window into it. - History shows on both sides. - Messages from either side stay in sync. - Approval requests appear on both the Mac and the other device. - Your Mac must stay on and awake. ## Connect a chat 1. In a Claude chat, run /remote-control, or click Claude Remote Control in the chat header and choose Enable. 2. Open the chat in a browser from the banner above the message box, or pick it up in the Claude mobile app. 3. Run the command again, or choose Disconnect, to stop. > **Needs a Claude plan sign-in** > > Sign in with claude auth login on a Pro, Max, Team or Enterprise plan. An API key does not work for remote control. --- # Project settings > Share setup, environment variables and run commands across every task in a code project. Source: https://videodraft.ai/docs/project-settings ## Where settings live Open Settings > Project settings to edit a project’s configuration. It is saved in a file at the project root, which you can also edit by hand. Project root: ``` /.videodraft/settings.toml ``` Settings stack in layers. A later layer overrides an earlier one. A project without a file keeps the defaults. | Layer | File | Use it for | | --- | --- | --- | | Your defaults | ~/.videodraft/settings.toml | Settings you want in every project. | | Project | /.videodraft/settings.toml | Shared settings, committed for the team. | | Personal | /.videodraft/settings.local.toml | Your own overrides. Kept out of Git. | ## Copy local files into new tasks A new isolated task starts from a clean copy of the repository, so files Git ignores, such as .env.local, are missing. Files to copy brings them across. - Enter one pattern per line. An empty box copies .env* files. - A .worktreeinclude file at the repository root takes priority over this setting. - ADE copies only files Git ignores. It skips symlinks, binary files, files over 5 MB and files that already exist. - The task’s welcome screen shows how many files were copied. > **Copying is not setup** > > Installing dependencies and other setup still follow your project’s own instructions or setup script. ## Environment variables and run commands - Environment variables set here reach the setup script, the Run tab and the agent. - Variables load when a chat starts. Run /restart after changing them. - Run commands appear in the Run button’s menu in the right panel. - Each task gets its own VIDEODRAFT_PORT, so tasks can run servers at the same time. --- # Git and pull requests > Give every task its own branch, review the diff and open a pull request without leaving ADE. Source: https://videodraft.ai/docs/git-and-pull-requests ## A branch for every task When you open a repository with Isolated workspaces, or clone one, each new task gets its own branch and its own worktree: a second copy of the repository on disk. Tasks can run at the same time without touching each other or your main checkout. - A new task branches from the repository’s default branch. - New task from branch starts from a remote branch or an open pull request instead. - The branch you start from is the task target. Changes and pull requests compare against it. - Change the target from the menu next to the branch name in the chat header. - A repository you chose to edit directly has no worktrees. Every task works in that folder. > **When the target can change** > > Not while the agent is running or a pull request is open. If the task has no changes yet, ADE moves it to the new target right away. ## Review before you ship The Changes tab shows the task’s diff against its target. Comment on it yourself, ask for an AI review or get a second opinion from Codex. Reviews run in their own chat. | Review | How | | --- | --- | | Your comments | Click + in the diff gutter next to a line or range, write in Add a comment for the AI and choose Comment. | | AI review | Click Review in the right panel header. A new chat reviews the diff and adds its findings to it. | | /codex-review | Gets an independent Codex review of the same changes, when Codex is connected. | | /codex-challenge | Asks Codex to argue against the approach. | | /security-review | Checks the branch’s pending changes for security problems. | > **Comments go with your next message** > > There is no separate send button. Pending comments attach to your next normal message in that chat, not to a slash command or a queued message. ## Checks and pull request status - The Checks section of the Git tab shows CI results for the branch. - When checks fail, choose Fix errors to send the failures to the agent. - Status rows show where the pull request stands, such as Waiting for PR review or PR changes requested. - On PR changes requested, Add comments to prompt gives the reviewer’s comments to the agent. ## Open and merge a pull request Pull request actions sit in the right panel header. They run through the GitHub CLI, so they use your gh sign-in and its permissions. | Action | What it does | | --- | --- | | Commit & Push | Stages everything, commits with your message and pushes. | | Push / Pull | Pushes to or pulls from the branch’s remote. | | Create PR | Opens a pull request against the task target. | | Create draft PR | Opens it as a draft. Mark it ready later from the same menu. | | Create PR manually | Opens GitHub’s pull request page in your browser. | | View PR | Replaces Create PR once a pull request is open. | | Merge | Squash-merges the open pull request. | | Auto-merge | Turns GitHub auto-merge on or off. | | Resolve conflicts | Asks the agent to resolve merge conflicts on the branch. | > **After the merge** > > The right panel offers Continue or Archive. Continue starts a fresh branch in the same task and keeps its chats. Archive closes the task. ## Answer review comments While a pull request is open, the Git tab lists review threads that still need an answer. Resolved and outdated threads are hidden. - Open a thread to jump to that comment on GitHub. - Add the threads to the chat, and the agent works through them. - The list refreshes when the agent finishes. ## What you need - A Git repository with a remote for push and pull request actions. - The GitHub CLI signed in. ADE asks once if it is not. - Code mode. VideoDraft mode hides the Git controls. While a chat in the task is running, ADE turns off actions that change the repository. This only affects that task. --- # Controls and shortcuts > Keyboard shortcuts, slash commands, and how to arrange tabs and windows in ADE. Source: https://videodraft.ai/docs/controls-and-shortcuts ## Main shortcuts | Shortcut | Action | | --- | --- | | Cmd + K | Open Command Center | | Cmd + / | Open Keyboard Shortcuts | | Cmd + , | Open Settings | | Cmd + O | Switch between VideoDraft mode and Code mode | | Cmd + P | Open Prompt Library | | Cmd + N | Start a new task | | Cmd + T | Open a new chat tab | | Cmd + W | Close the current tab | | Cmd + Shift + F | Open History, your recent chats across every project | | Cmd + [ / Cmd + ] | Go back or forward | | Cmd + Shift + [ / Cmd + Shift + ] | Previous or next tab | | Ctrl + Shift + Tab / Ctrl + Tab | Previous or next tab | | Cmd + Shift + N | Open the current chat in its own window | | Cmd + \ | Split the center into side-by-side tabs | | Cmd + Opt + Left / Right | Move to the previous or next pane | | Cmd + L | Focus the message box | | Cmd + D | Start or stop dictation | | Cmd + U | Add an attachment | | Opt + T | Change the thinking level | | Cmd + F | Find text in the chat | | Cmd + Opt + T | Switch between light and dark | ## Arrange tabs and windows | To do this | Do this | | --- | --- | | See two tabs side by side | Press Cmd + \, or right-click a tab and choose Split right. | | Move a tab to the other pane | Right-click the tab and choose Move to next pane. | | Keep a previewed file open | A preview tab has an italic title and the next file replaces it. Double-click the tab, or choose Keep open. | | Give a chat its own window | Right-click the tab and choose Open in new window, or press Cmd + Shift + N. | ## Slash commands Type / in the message box to see every command for the current agent and mode. These are the ones you will use most. | Any agent | Action | | --- | --- | | /clear | Start a fresh chat. | | /plan | Turn Plan mode on or off. | | /model | Switch the model. | | /compact | Summarize the chat to make room. | | /restart | Restart the agent to pick up setting changes. | | /add-dir | Link another task or any folder. | | Claude chats | Action | | --- | --- | | /remote-control | Continue this chat from your phone or browser. Run it again to disconnect. | | /ask-codex | Hand a task to Codex. Needs Codex connected. | | /codex-resume | Continue the latest Codex thread in this task. | | /codex-review | Get a Codex review of this task’s changes. Code mode only. | | /codex-challenge | Have Codex challenge the approach. Code mode only. | | Codex chats | Action | | --- | --- | | /goal | Set a goal for a long-running task. Codex keeps working until it is met. | | /goal pause | Pause the goal. | | /goal resume | Resume the goal. | | /goal clear | Clear the goal. | --- # Privacy and permissions > What stays on your Mac, what leaves it, and what an agent can reach. Source: https://videodraft.ai/docs/privacy-and-permissions ## What stays local and what leaves - ADE, your chat history, tasks and terminals run on your Mac. - VideoDraft Editor projects, edits, previews and exports stay on your Mac. - To answer, ADE sends your chat and the files the agent reads to the company behind the agent you chose. - VideoDraft media tools upload your prompt and input media to VideoDraft servers. - Dictation sends microphone audio to the service picked in Settings > Dictation > Transcription provider: Gemini Smart or Deepgram. ## File access follows the task An agent starts in the current task. Attachments and linked folders widen what it can read. A folder you edit directly is changed in place, so pick it carefully. - Check which task is active before asking for anything destructive. - Link only the folders a task needs. - Use Plan mode to see an approach before files change. - Give scheduled tasks fewer folders and smaller jobs. Nobody watches them run. Plan, Ask and Auto decide when an agent stops for your approval; see [Work with agents](https://videodraft.ai/docs/working-with-agents#controls). When an agent asks, read the command or file path on the card, then allow or deny it. ## Sign-ins ADE uses the Claude Code and Codex sign-ins already saved on your Mac. Signing in to VideoDraft creates a separate sign-in for ADE and does not change your own videodraft CLI setup. ## macOS permissions | Permission | When VideoDraft asks | | --- | --- | | Notifications | Optional. Alerts you when a background reply finishes or an agent needs you. | | Microphone | Optional. Asked when you first use dictation. | | Screen Recording | Optional. Needed to capture a browser preview screenshot into a chat. Restart ADE after granting it. Computer Use asks for its own access in Settings > Computer use. | | Accessibility and Background input | Optional. Needed only for Computer Use. macOS lists the helper as VideoDraft Computer Use. | ## Enterprise data privacy Your company can turn on Enterprise data privacy for everyone on its VideoDraft account. Settings > Account then shows Enterprise data privacy, marked Managed. - Features that rely on outside AI services turn off, such as AI chat titles, custom MCP servers and importing custom agent settings. - VD Code, Grok Build, Computer Use, dictation and telemetry are unavailable. - Agents can still read your other chats and diffs, because that runs on your Mac. --- # Fix a problem > Find your problem, take the first step, and know what each recovery action keeps. Source: https://videodraft.ai/docs/troubleshooting ## Start with the problem | Problem | First step | | --- | --- | | An agent will not connect | Refresh it in Settings > Providers. [Details](https://videodraft.ai/docs/troubleshooting#provider). | | A reply was cut off | Check the files, then click Continue on that turn. | | A reply failed | Read the error, fix the cause, then click Retry. | | You changed agent settings, or the agent stopped responding | Type /restart. | | The chat is very long | Ask for a handoff note, then /compact or start a new chat. | | You want to try another direction | Fork the chat, or start a new task. | | An agent’s file changes were wrong | Use [Reset to this point](https://videodraft.ai/docs/troubleshooting#reset). | | A timeline edit was wrong | Use Undo in VideoDraft Editor. | | A generation is still running | Check its status. Do not submit it again. | | Media is missing | See [missing media](https://videodraft.ai/docs/troubleshooting#media). | ## An agent will not connect 1. Open Settings > Providers and refresh the agent. 2. Use its sign-in button, or run claude auth login or codex login in Terminal. 3. Return to ADE and refresh again. 4. If an agent download fails, click Retry now on the banner in the chat. ## What Reset to this point can undo Reset rolls the task’s files back to how they were at that turn and removes the later turns. It is not offered on the latest turn. In a folder without Git there is nothing to roll back to, and ADE tells you. - Reset restores files Git tracks. Media files Git does not track stay as they are. - It does not undo cloud generations, charges or anything the agent did outside your Mac. - After a reset, some agents start a fresh conversation and do not remember the earlier turns. - Later chat history is removed. Read the confirmation first. ## Media is missing or moved 1. Look in the task’s media/ folder in All files. 2. Ask the agent for the exact saved path. 3. Check the generation finished and did not stop on an error. 4. For media made in the web app, open the right AI Studio project or video project. 5. If a linked .vdproject moved, the Editor tab shows Project not found. Click Link a project… and pick its new location. ## A scheduled task did not run - Check the scheduled task is on and shows a next run time. - Open its run history and look for skipped runs, stopped runs or sign-in errors. - Check [ADE was open and awake](https://videodraft.ai/docs/scheduled-tasks#behavior) at run time. ADE cannot wake a Mac that is shut down. ## VideoDraft mode or the editor is unavailable - Check your Mac has Apple silicon and macOS 15 or later. - Open Command Center (Cmd + K) and choose Open VideoDraft Editor to show the window. - If the Editor tab says No project linked yet, click Create project now or Link existing project…. - A closed editor window can still be running for agents. ## Browser screenshots do not work 1. Open Settings > Experimental, turn on In-app browser preview and find Screen Recording below it. 2. Allow VideoDraft in System Settings > Privacy & Security > Screen Recording. 3. Quit and reopen ADE. macOS applies this permission only after a restart. 4. Try the screenshot again. ## Keep work safe and get help Back up important task files, media and editor projects. Chat history, Git checkpoints, editor autosave and web history each save different things. None of them is a full backup. When you contact support, include the app version, agent, what you did, the exact error and whether files or jobs changed. Leave keys and passwords out of screenshots. - [Contact support](mailto:contact@videodraft.ai) --- # VideoDraft Editor > A video editor on your Mac that you and your agents work in together. Source: https://videodraft.ai/docs/native-editor ## One editor, two ways to work VideoDraft Editor is a Mac app that comes with ADE. It runs on your Mac, not in the cloud. You and your agents edit the same project, with the same undo and the same export. - Agents can edit with the window closed. Click Open in Editor, or use Command Center, to watch or edit by hand. - Changes an agent makes show up in the open window as they happen. - Closing the window does not stop an agent’s edit. - Projects are .vdproject packages and open without ADE. ![VideoDraft Editor with the Library, preview, Inspector and timeline.](https://videodraft.ai/ade-screenshots/film-editor.webp) The Library is on the left, the preview in the middle, the Inspector on the right and the timeline below. > **What you need** > > An Apple silicon Mac with macOS 15 or later. ## Start or link a project 1. In ADE’s right panel, open the Editor tab. 2. Click Create project now, or Link existing project… to use a .vdproject you already have. You can also ask the agent to make one. 3. Click Open in Editor to see the window. 4. Import files with File > Import Media… (Cmd + I), or ask the agent to place them. 5. Edit by hand, ask the agent, or both. 6. Export with File > Export… (Cmd + E), or ask the agent. ## By hand or with an agent Most editing works both ways. Some jobs need an agent for now. | Job | By hand | With an agent | | --- | --- | --- | | Cut, trim and arrange clips | Yes | Yes | | Speed changes and ramps | Yes, in the Inspector or Modify > Retime | Yes | | Transitions | Yes, from Transitions in the Library, or Cmd + T | Yes | | Remove dead air | One quiet stretch at a time: right-click it on the audio clip and choose Remove Dead Air | All at once | | Cut by words, remove filler words | Not yet | Yes | | Make captions | Restyle them in Inspector > Text | Yes | | Match color between shots | Adjust in Inspector > Color | Yes. The agent measures the shots and matches them. | | Split screen and picture-in-picture | For multicam clips: right-click and choose Layout | For any clips | ## Captions Agents make captions from the speech on the timeline, transcribed on your Mac. When you ask, say how many words per caption and where they should sit. Restyle them anytime in Inspector > Text. ## Color, audio and effects These stay editable, so you or an agent can change them later without making new media. | Area | What it covers | | --- | --- | | Color | Exposure, contrast, curves, color wheels, hue controls and LUTs (ready-made color looks). | | Audio | Volume, volume keyframes (Cmd + click an audio clip), fades, dead air removal and Preserve pitch on sped-up clips. | | Effects | Blur, grain, glow, vignette, chroma key (green screen removal), video noise reduction, masks and blend modes. | | Transitions | Cross Dissolve, Dip to Black, Dip to White, Blur Dissolve, Push, Linear Wipe, Whip Pan and Cross Zoom. | | Multicam | Sync several camera angles, switch between them and show them side by side. | ## Export and hand off Choose File > Export… (Cmd + E), then pick an output. | Output | Use it for | | --- | --- | | Video | H.264, H.265, HEVC 10-bit HDR or ProRes, at 720p, 1080p, 2K, 4K or Match Timeline. | | Timeline: XMEML | Sending the edit to Premiere Pro, or DaVinci Resolve through its XML import. | | Timeline: FCPXML | Sending the edit to Final Cut Pro or DaVinci Resolve. Includes titles. | | Project package | A self-contained copy of the .vdproject and its media. | > **Timeline exports carry the cut, not the look** > > XMEML and FCPXML carry cuts, speed changes, audio, position, crop, opacity and keyframes, but not color or effects. XMEML also turns a speed ramp into one average speed; FCPXML keeps the ramp. Export a finished video when the look must survive. --- # A narrated video, start to finish > Turn a voiceover and a few clips into a finished video with captions, an end title and music. Source: https://videodraft.ai/docs/editor-walkthrough ## What you need - A short script, 15 to 30 seconds when read aloud. The example below works. - A voiceover of that script. Record it on your Mac with QuickTime Player > File > New Audio Recording, or ask the agent for a practice voiceover, which uses media credits. - Four or five clips or images to cover the words. - A music track you have the rights to use. Optional. script.txt: ``` Every team says they move fast. Then the handoffs start. A brief here. Feedback there. Three tools before anyone sees a cut. Northwind puts it in one place. Write, review, and ship from the same board. Northwind. Ship the work, not the process. ``` Put the script in your task folder and the recording, clips and music in media/. Most steps work by hand or with an agent; the ones that need an agent include a prompt. > **Cut the voice first** > > When narration leads a video, get its timing right before placing any shots. The pictures then follow the words. ## Create the project Open VideoDraft Editor with Open in Editor in ADE, or from Command Center. Choose File > New Project (Cmd + N). In Create project, set the name, format and Save Location. | Setting | How to choose | | --- | --- | | Aspect Ratio | Widescreen 16:9 for web and YouTube. Vertical 9:16 for Reels, TikTok and Shorts. Square 1:1 for feed posts. Cinema 2.4:1 for a wide look. | | Resolution | 1080p suits most videos. Choose 1440p or 4K if you plan to zoom in or reframe. Custom sets any size. | | Frame Rate | Match your footage. 24 fps looks like film, 30 fps is the safe web choice, 60 fps holds fast motion. | If you are unsure, choose Widescreen 16:9, 1080p and 30 fps. The editor saves as you work. Still keep backups of the project and its media. ## Bring in the media Choose File > Import Media… (Cmd + I), or drag files onto the Library on the left. Import the voiceover, the clips and the music. > **No final voiceover yet?** > > Ask the agent for a practice voiceover from the script and time the edit to it. Swap in the real recording when you have it. ## Lay down the voiceover Drag the voiceover onto an audio track at the very start. Its length is now the length of your video. Tracks are layers. Video, images and titles go on video tracks; audio goes on audio tracks. A video clip with sound brings its audio along, and the two stay in sync when you trim or change speed. ## Tighten the voiceover 1. Remove dead air, the quiet gaps with no speech. The editor marks them on the audio clip. By hand, right-click one and choose Remove Dead Air. To remove them all, ask the agent. 2. Ask the agent to remove filler words or a fumbled phrase. 3. When you ask, say how tight the cuts should be: tight, balanced or loose. Ask the agent: ``` Remove the dead air from the voiceover and any filler words. Use balanced cuts, then tell me the new length. ``` > **Quiet music can look like dead air** > > A soft intro or fade in music can be marked as dead air. Check the result before removing it from a music track. ## Cover the voice with shots Put the clips on a video track above the voiceover. Go line by line: each sentence gets a shot. Press Cmd + B to cut a clip at the playhead, and Opt + [ or Opt + ] to trim its start or end. See [all editor shortcuts](https://videodraft.ai/docs/editor-guide#editing). - A shot is too long: speed it up in the Inspector, or add a ramp so the speed changes across the clip. - Two shots need to blend: add a transition from Transitions in the Library. It belongs to the second clip. - Two shots should share the screen: ask the agent for a split screen or picture-in-picture. ## Captions and the end title Ask the agent for captions: ``` Make captions from the voiceover, 4 words each, low in the frame, highlighting the word being spoken. ``` | Option | Where to change it | | --- | --- | | Words per caption | In your request. 3 to 5 words reads well on a phone. | | Highlight color | Inspector > Text, Highlight. | | Animation | Inspector > Text. Line animation moves the whole caption; Word animation moves each word. | | Position | Inspector > Text, Position. Low in the frame, clear of the edge. | | Letter case and profanity masking | In your request. | For the end title, add text from Text in the Library over the last few seconds and type your website or call to action. Style it in Inspector > Text. ## Color, music and export 1. Match the shots’ color in Inspector > Color, or ask the agent to match them to one shot. 2. Put the music on an audio track under the voice. Cmd + click the music clip to add volume keyframes, lower it while the voice speaks and raise it in the gaps. 3. Watch the whole thing once without touching anything. 4. Choose File > Export…, pick Video, H.264 and 1080p, and export. - [Export formats](https://videodraft.ai/docs/native-editor#export) - [Editor window and shortcuts](https://videodraft.ai/docs/editor-guide) ## Or hand it all to an agent An agent works in the same project with the same undo, so you can hand over the whole job or take over at any point. One prompt for the whole job: ``` Build a 16:9, 1080p, 30 fps video from script.txt and the media in this folder. Put the voiceover on an audio track, remove dead air and filler words. Cover each line of the script with one of the clips, cutting on the voice. Make captions, 4 words each, low in the frame, with the spoken word highlighted. Add [your website] as text for the last four seconds. Match the shots’ color, put the music under the voice and lower it while the voice speaks. Export H.264 1080p and show me the file. ``` > **Watch it happen** > > Open the editor while the agent works. Changes appear live, and you can take over whenever you like. --- # Editor window and shortcuts > The four areas of VideoDraft Editor and the keys that matter. Source: https://videodraft.ai/docs/editor-guide ## The four areas ![VideoDraft Editor with a vertical video in the preview, the Inspector open and a timeline of shots, captions, dialogue, sound effects and music.](https://videodraft.ai/ade-screenshots/home-editor.webp) The Library on the left, the preview in the middle, the Inspector on the right and the timeline along the bottom. | Area | What it holds | | --- | --- | | Library, left | Media, Colors, Text, Effects and Transitions. Import with Cmd + I, sort into folders and search. Colors appears after View > Show in Workspace > Colors Palette. | | Preview, center | Plays the timeline. Drag a clip to move it and drag a corner to resize it. Shift + T turns on Transform and Shift + C turns on Crop. | | Inspector, right | Settings for what is selected, in tabs: Video, Color, Effects, Animate, Audio, Text, Multicam and Info. With nothing selected, it shows the project settings. | | Timeline, bottom | Where you edit. Tracks are layers: the top video track covers the ones below it. | | Key | Action | | --- | --- | | Ctrl + Cmd + 1 | Show or hide the Library | | Cmd + 4 | Show or hide the Inspector | | Ctrl + Cmd + F | Full screen | | View > Maximize Panel | Fill the window with one area. Esc returns. | | + / - or Cmd + = / Cmd + - | Zoom the timeline | | Shift + Z | Fit the whole timeline | | Opt + scroll, or pinch | Zoom at the pointer | | Cmd + scroll | Scroll the timeline sideways. Over the preview, it zooms the preview. | Drag a number field in the Inspector to change it, or type a value. Help > Keyboard Shortcuts (Cmd + /) lists every shortcut. ## Moving around | Key | Action | | --- | --- | | Space | Play or pause | | J / K / L | Play backward, stop, play forward | | Left / Right | Step one frame | | Shift + Left / Right | Skip backward or forward | | Up / Down | Previous or next edit | | Home / End | Go to the start or end | | I / O | Set range start or end | | Opt + X | Clear the range | Shift + drag along the ruler to select a range. ## Editing | Key | Action | | --- | --- | | A | Select tool | | B | Blade tool | | R / Z / H | Range Selection, Zoom and Hand tools | | Cmd + B | Blade at the playhead | | Opt + [ | Trim the start to the playhead | | Opt + ] | Trim the end to the playhead | | Backspace | Delete and close the gap | | Fn + Delete | Replace with Gap | | Cmd + T | Add Cross Dissolve | | N | Snapping on or off | | Opt + Left / Right, or , / . | Nudge the selection one frame | | Esc | Deselect and go back to the Select tool | | Ctrl + Shift + S | Detach Audio | | Cmd + G / Cmd + Shift + G | Nest Clips / Break Apart Nest | Backspace moves every later clip earlier. Fn + Delete leaves an empty space instead, for a shot you have not placed yet. To select everything after a point, use Edit > Select > Clips Ahead on Track. | With the mouse | Action | | --- | --- | | Shift + drag an edge | Ripple trim: trim and move the later clips with it | | Cmd + drag media in | Insert: drop it in and push later clips along | | Opt + drag a clip | Duplicate it | | Opt + click or drag | Work on a clip without its linked audio or video | | Cmd + click an audio clip | Add a volume keyframe | Clips snap to each other and to the playhead. A trimmed edge can always be dragged back out to the clip’s full length. ## Speed and ramps Select a clip and open its speed settings in the Inspector: Playback on the Video tab, or Timing on the Audio tab. Set a Rate from 0.01x to 1000x, or set Mode to Curve and choose a Ramp preset so the speed changes across the clip. You can also right-click a clip and choose Speed. Animation and volume changes stretch with the clip, and its sound speeds up or slows down too. Preserve pitch, under More, keeps voices sounding natural. | Key | Action | | --- | --- | | Cmd + Opt + 1 to 5 | Slow 0.25x, Slow 0.5x, Normal 1x, Fast 2x, Fast 4x | | Cmd + Opt + 0 | Reset Speed | | Cmd + Opt + R | Ripple Later Clips on or off | > **Speed changes a clip’s length** > > If a slower clip would run into the next one, the editor refuses the change. Move the next clip first, or turn on Ripple Later Clips. ## Multicam Multicam is one scene filmed by two or more cameras. Ask an agent to group the camera files. The editor syncs them by timecode, or by their sound, and you choose which angle shows at each moment. To show several angles at once, right-click a multicam clip and choose Layout. You can pick side by side, top and bottom, picture-in-picture, a 2×2 grid, main plus sidebar, or three-up. Clip > Ungroup Multicam separates the group again. ## What this version cannot do yet - Dead air detection skips recordings longer than 30 minutes. - You cannot play a clip in reverse, and there is no smoothing for slow motion. - There is no audio noise reduction. Video noise reduction is available as an effect. --- # VideoDraft on the web > Make images, video and sound, or take a video from idea to finished cut, in your browser. Source: https://videodraft.ai/docs/web-app ## Where to start | You have | Go to | | --- | --- | | An idea for one image, clip or sound | AI Studio in the sidebar. | | An idea, script or voiceover for a whole video | Home. Type or paste it in the box and pick Idea, Script, Voiceover, Storyboard or Short Film. | | A script you want to write yourself | Script, then From Scratch. | | A character or product that must look the same in every shot | Character Studio or Product Studio. | | A presenter who talks to camera | Avatar Video. | | Your own voice to use in videos | Voice Cloning. | | Questions or changes for a video project | VideoDraft Agent. | - [Your first image and clip](https://videodraft.ai/docs/web-quickstart) - [Scripts and storyboards](https://videodraft.ai/docs/web-storyboards) ## Projects in the web app A video project holds a script, storyboard and timeline. Find them under Projects in the sidebar. An AI Studio project is a collection of images, video and sound you made. When you open AI Studio from inside a video project, character or product, the results are saved there. Everything here lives in the cloud. To use media on your Mac, download it. - [AI Studio basics](https://videodraft.ai/docs/web-ai-studio) - [How credits work](https://videodraft.ai/docs/providers-and-usage?mode=web) --- # Your first image and clip > Make a product photo in AI Studio, create a variation, turn it into a short clip and download it. Source: https://videodraft.ai/docs/web-quickstart ## Set up a project You need a VideoDraft account with credits. This tutorial makes two images and one short clip; the Create button shows each cost before you click. - [Open AI Studio](https://app.videodraft.ai/ai-studio) - [How credits work](https://videodraft.ai/docs/providers-and-usage?mode=web) 1. Click AI Studio in the sidebar. 2. Click + (New Project) above the Projects list. Name it First product shot and click Create. 3. Click the Image Studio icon. Click the model button and pick an image model. 4. Set Aspect Ratio to 16:9 and Count to 1. ## Make one image Image prompt: ``` A small matte blue ceramic coffee cup on a warm oak desk beside a window. Soft morning light from the left, eye-level product photograph, uncluttered background, no writing or logo. ``` 1. Paste the prompt into the prompt box. 2. Check the cost on the Create button, then click Create. 3. When the image appears, click it to open the preview. Check the shape, edges, light and layout. > **Still working?** > > Leave it running. You can close the page and come back to the same project. Clicking Create again starts, and pays for, a second image. ## Make a variation 1. Hover the image and click Reference. It becomes a reference for the next image. 2. Replace the prompt with the one below. 3. Optional: click Enhance Prompt to have the prompt rewritten for the chosen model. Read the new text before you click Create. 4. Click Create. Compare the two and click the star on the one you like. Variation prompt: ``` Keep the same cup, light and desk. Move the cup slightly to the right and leave clear space on the left for a title. ``` ## Turn it into a clip 1. Hover your starred image and click Video. Video Studio opens with the image attached. 2. Pick a video model, then a Duration and Resolution. Check the image is still attached. 3. Paste the motion prompt below, check the cost and click Create once. 4. Play the finished clip, with sound if Audio is on. Motion prompt: ``` A slow, steady push toward the cup. The cup stays still. Morning light stays the same. No new objects or text. ``` ## Download the result 1. Star the clip too. On larger screens, the Favorites button at the top shows only starred items. 2. Hover the clip and click Download. The Download menu at the top can also download All, Favorites, Images or Videos. 3. Open the downloaded file to check it plays. 4. Come back to First product shot in the Projects list anytime. - [AI Studio basics](https://videodraft.ai/docs/web-ai-studio) - [Make video](https://videodraft.ai/docs/web-video-generation) --- # AI Studio basics > Find your way around AI Studio, pick models, change results and keep projects organized. Source: https://videodraft.ai/docs/web-ai-studio ## The screen | Part | What it does | | --- | --- | | Projects list | Your AI Studio projects. + makes a new one, Search finds old ones and pinned projects stay at the top. | | Image Studio, Video Studio, Sound Studio | The three icons that switch what you make. Each has its own models and settings. | | Model button | Shows the current model. Click it to search and pick another. | | Prompt box | Your prompt, references and Enhance Prompt. | | Settings | Aspect Ratio, Count, Duration, Resolution, Quality and Audio, depending on the model. | | Create | Starts the job. It shows the cost, such as “12 cr”. | | Gallery | Your results. Failed ones show Failed with a Retry button. | | Prompt Engineer | The round button at the bottom right. Chat about a prompt before you make anything. | ## Pick a model | You want to | Pick a model that offers | | --- | --- | | Make a new image | Text to image, in the aspect ratio you need. | | Keep a person or product the same, or change an image | Image references or editing. | | Make video from an image or footage | See [how inputs go in](https://videodraft.ai/docs/web-video-generation#mode). | | Make speech, music or effects | Sound Studio, with the right Category. | | Improve a result you like | Upscale, or an interpolation model, which adds in-between frames for smoother motion. | - Settings change with the model. After switching, check the ratio, resolution and audio again. - Cost depends on the settings, not just the model. Watch the number on Create. > **Models change** > > The model button and the Create cost always show what is available now and what it costs. ## Do more with a result Hover a result for quick actions, or open its … menu for all of them. | Action | What it does | | --- | --- | | Star | Adds it to Favorites. | | Download | Saves the original file. | | Reference | Uses this result to guide your next image or video. In the menu: Use as Reference. | | Video | Opens Video Studio with this image attached. In the menu: Create Video. | | Edit | Makes a changed version of this result. In the menu: Continue with Edit. | | Continue with Extend | Makes a video longer. | | Upscale | Makes a sharper, larger version. It costs credits. | | Recreate | Puts this result’s prompt and settings back in the prompt box. Add references again. | | Optimize Prompt | Sends the prompt to Prompt Engineer to improve it. | | Copy Prompt | Copies the prompt. | | Delete | Removes the result. | ## Change one thing in an image 1. Hover the image and click Edit. 2. Say what must stay the same and the one thing to change. 3. Check the image is attached, then click Create and compare the two. Example edit: ``` Keep the cup’s shape, the camera angle and the light. Change only the background to a pale warm gray. Keep the shadow under the cup. ``` To see the same scene from another angle, open the image in the preview and choose Change Camera Angle. When the content is right, click Upscale for a sharper version. Upscaling cannot fix a wrong face, logo or object, so fix those first. ## Enhance Prompt or Prompt Engineer Enhance Prompt rewrites the text in the prompt box for the model you picked. Read the new version before clicking Create. Prompt Engineer is a conversation: ask it for ideas, explain your goal or have it improve a prompt from a result. ## Keep projects organized Open a project’s … menu in the Projects list. | Menu item | What it does | | --- | --- | | Pin | Keeps the project at the top. | | Rename | Changes its name. | | Settings | Opens Project Settings. | | Assets & Context | Add adds files and context. Copy to new project starts a new project with them. | | Share | Shares the project. See [sharing](https://videodraft.ai/docs/web-sharing). | | Delete | Deletes the project. | Project Settings sets defaults for Image, Video and Sounds, such as the model, aspect ratio and voice. Click Save after changing them. Its Usage view shows what the project has spent. --- # Make video > Choose how images and video go in, direct the motion and check the finished clip. Source: https://videodraft.ai/docs/web-video-generation ## How your inputs go in Each video model shows its own options above the prompt box, such as Frames, References, Edit or Extend. | You have | Do this | | --- | --- | | Only an idea | Describe the shot. Good when the exact first frame does not matter. | | An image to start from | Hover the image and click Video, or choose Create Video from its menu. | | A start and an end image | Pick a model with start and end frame slots. Swap start and end frame flips them. | | References for a look or character | Pick a model with a References option and add them there. | | A video whose movement you want to copy | Pick a Motion Control model. Add the character image and the motion video. | | A clip to change or lengthen | Hover it and click Edit Video, or choose Continue with Extend from its menu. | Models use an image in different ways. Some start the clip on that exact frame; others use it as a reference for the look. ## Make a controlled shot 1. Click the Video Studio icon and pick a model. Add the images or video it needs. 2. Set Duration, Resolution, Aspect Ratio and Audio. 3. Describe one clear action and one camera move. 4. Check the cost on Create and click it once. 5. Play the whole clip, with sound. Image to video: ``` Keep the same cup and desk. The camera slowly pushes forward. Morning light stays steady, no shake. The cup does not move and nothing new appears. ``` ## Improve or download Choose Upscale Video in the preview for a sharper version. For smoother motion, pick an interpolation model, which adds in-between frames. Each costs credits. Then download the clip or use it in a video project. > **A failed clip** > > Failed clips show Failed with a Retry button. Read the error and fix the cause first, such as a missing image or an unsupported length. - [Fix a problem](https://videodraft.ai/docs/web-troubleshooting) - [Bring clips into ADE](https://videodraft.ai/docs/web-to-ade) --- # Speech, music and sound > Make narration, dialogue, music and sound effects, change a voice, dub a clip or clone your own voice. Source: https://videodraft.ai/docs/web-sound ## Pick a category In AI Studio, click the Sound Studio icon and choose a Category. | Category | You give it | | --- | --- | | Voice | Text and a voice. | | Seed Audio | A prompt, a voice and optional reference audio or images. More control over the performance; open Settings > Advanced for the options. | | Dialogue | Lines of text, each with a voice. | | Music | A description of the music, plus length or lyrics where offered. | | Sound Effects | A description of the sound. | | Voice Changer | A recording and the voice to turn it into. | | Dubbing | Audio or video and the language to dub into. | ## Make a narration 1. Set Category to Voice and pick a voice from the Voice list. 2. Type a short test line, click Create and listen. 3. Fix pronunciation or pace in the text, then make the full read. 4. Download it and listen to the whole file. Test line: ``` Meet Northwind. Bring the brief, the references and the finished work together. ``` > **Voices from your own ElevenLabs account** > > They need your ElevenLabs key. See [using your own keys](https://videodraft.ai/docs/providers-and-usage#byok). ## Make a dialogue 1. Set Category to Dialogue and add a line for each speaker. 2. Pick a voice for each line and type its text. 3. Make a short exchange first and check who says what, the pauses and pronunciation. 4. Then make the full scene. ## Make music or a sound effect Music: ``` A calm instrumental bed for a product reveal: warm keys, soft percussion, steady pace, no vocals. Leave room for narration. ``` Sound effect: ``` A soft ceramic cup set down on a wooden desk. Close up, quiet room, no voice or music. ``` Match the length to your edit. Make each sound effect on its own so you can place it exactly. ## Change a voice or dub a clip 1. Set Category to Voice Changer or Dubbing and upload the file. 2. For Voice Changer, pick the new voice. For Dubbing, pick the language. 3. Check the cost on Create and click it. 4. Listen to the whole result next to the original. Dubbing changes the audio, not the mouth movements on screen. ## Clone your own voice 1. Click Voice Cloning in the sidebar, then Clone New Voice. 2. Choose Record Voice and read the sample text aloud, or Upload File. 3. Tick I allow VideoDraft to clone my voice and use it for my projects. 4. Name the voice, add a description and click Clone Voice. 5. Click Use Voice and make a short test line before using it everywhere. - Only clone your own voice, or one you have permission to use. - Record in a quiet room, at a steady distance from the microphone, with no music or echo. - [Make an avatar video](https://videodraft.ai/docs/web-avatar-video) - [How credits work](https://videodraft.ai/docs/providers-and-usage?mode=web) --- # Scripts and storyboards > Write the script, turn scenes into shots, and check the timing before you make final video. Source: https://videodraft.ai/docs/web-storyboards ## Start with a script 1. Click Script in the sidebar. 2. In Create Script, choose From Scratch to write it yourself, or AI / Script to start from an idea. 3. AI / Script opens the box on Home. Pick Idea, Script, Voiceover, Storyboard or Short Film, then describe the video or paste your script. 4. Read the scenes it makes. Ask VideoDraft Agent to rewrite any part that needs it. Example brief: ``` A 15-second product promo for a blue ceramic cup. Audience: home coffee drinkers. Three shots: a morning desk, a close-up of the cup and a calm final hero shot. Calm, clear narration. End with one short call to action. ``` - Say what viewers should understand or do at the end. - Read the narration aloud to judge the length. - Keep names, product details and pronunciation the same throughout. ## Scenes hold shots A scene is a part of the story. Its shots are what the viewer sees. When you change one, check the action, framing, dialogue and references still agree. | Shot detail | Question it answers | | --- | --- | | Subject and action | What happens in this image? | | Framing | Wide, medium or close? What must be readable? | | Camera | Where is the camera and how does it move? | | References | Which character, product or place appears? | | Length and sound | How long is the moment and what do we hear? | ## Make the shot images 1. Open the storyboard from Projects, or click Storyboard in the sidebar. A new one asks for a Board name. 2. Read each shot’s description and attach the character or product references you chose. 3. Make one shot’s image first with Generate Shot Images to test the look. 4. When the look is right, click Generate All Shot Images, then Generate All. 5. Look at all the shots together. If one is wrong, remake just that shot. Export > Export PDF shares the board, and Export > Download Media saves the images. ## Check the timing with an animatic An animatic plays the storyboard images in order with sound. It shows whether the story is clear and each shot is the right length, before you pay for final video. It can also be a finished piece on its own. 1. Click Produce on the storyboard. 2. In Produce Video, choose Animatic / Story Reel. 3. Choose With Voiceover or Without Voiceover and set Shot Duration. 4. Click Produce and watch the result. Adjust shot lengths around the narration. - [Produce the final video](https://videodraft.ai/docs/web-production) - [Character Studio](https://videodraft.ai/docs/web-character-studio) --- # Produce and export a video > Turn a storyboard into a video, fix any failed scenes and export. Source: https://videodraft.ai/docs/web-production ## Before you produce Start with a short script and storyboard you are happy with. Check references, dialogue and length first. AI Production is on the Pro plan and above, and Produce Video shows an estimate before it spends credits. - [Scripts and storyboards](https://videodraft.ai/docs/web-storyboards) - [How credits work](https://videodraft.ai/docs/providers-and-usage?mode=web) ## Pick how to produce | Option | Use it when | | --- | --- | | Animatic / Story Reel | You want to judge timing with the storyboard images and audio. | | AI Production | You want VideoDraft to make a video for every scene. | | Make shots yourself in AI Studio | You want a specific model or tight control over each shot, then place the clips yourself. | > **AI Production picks its own models** > > It does not use every AI Studio model. If you need a particular one, make those shots in AI Studio and add them to the timeline. ## Produce the scenes 1. Open the storyboard and click Produce. 2. In Produce Video, choose AI Production. 3. Leave Generate videos now on to make every scene’s video right away. Turn it off to set up each scene first and make the videos later. 4. Check the estimate and click Produce. 5. Watch each scene as it finishes. AI Production writes its own narration for each scene. If your video must use a voiceover you recorded, make an animatic with it instead, or make the shots in AI Studio. ## Fix failed scenes 1. Open the failed scene and read its error. 2. Fix the cause, such as a missing reference or unsupported setting. 3. Remake just that scene. Do not produce the whole video again for one failure. 4. Replace any clip you do not like and play the cuts on either side of it. ## Finish and export 1. Open the timeline and check the order, trims, narration, captions and music. 2. Play it through and fix any clip that is missing or still processing. 3. Click Export, render the video and wait for it to finish. 4. Download the file and watch it to the end. To finish the edit in another app, choose Export > Export XML: FCPXML for Final Cut Pro, or XML for Premiere Pro and DaVinci Resolve. It carries the edit, not a rendered video. Free plans see Upgrade to unlock. ## Get help from VideoDraft Agent VideoDraft Agent in the sidebar can work on your video project with you. Name the scene, the change you want and what must stay the same, and check the result. | Part of the project | Ask for | | --- | --- | | Script | Scene structure, dialogue, narration or a rewrite of one scene. | | Storyboard | Shot framing, references and consistency between shots. | | Animatic | Timing, pacing and how audio lines up. | | Timeline | A change to order, length, captions or sound. | Example: ``` Read this project’s script and storyboard. Pick the three shots that matter most to the story. Propose a small first production pass with its cost, and wait for my choice before making anything. ``` - [Share the project](https://videodraft.ai/docs/web-sharing) - [Edit on your Mac](https://videodraft.ai/docs/web-to-ade) --- # Character Studio > Create a character once and reuse the same look in every shot. Source: https://videodraft.ai/docs/web-character-studio ## Create a character 1. Click Character Studio in the sidebar, then Create New Character. 2. Choose Upload Image to use a photo you have the rights to, or Design with AI to describe one. 3. With Design with AI, describe the face, clothing and details that must never change. Pick one under Choose Your Favorite, or click Regenerate. 4. Click Create Character. A clear face and simple background make the character easier to reuse. ## Add looks Open the character’s menu and choose View Looks to see and add looks, such as a full-body view or a different outfit. - Start with a clear front view. - Add a full-body or side view when your shots need it. - Say clearly when clothing or age should change. - Compare the looks side by side before using them. Look prompt: ``` Keep this character’s face, blue coat and red scarf. Full body, facing the camera, same style and light. Plain background, no props. ``` ## Use the character Choose Use in Studio from the character’s menu to make images and video with it in AI Studio. In a storyboard, attach the character to the shots it appears in before making their images. - [Make video](https://videodraft.ai/docs/web-video-generation) - [Share your work](https://videodraft.ai/docs/web-sharing) --- # Product Studio > Save a product once and keep its shape, packaging and materials right in every shot. Source: https://videodraft.ai/docs/web-product-studio ## Add a product 1. Click Product Studio in the sidebar, then Create New Product. 2. Choose Upload Images to use photos of the product, or Design with AI. 3. Fill in Brand, Description, Color and Material so the details stay accurate. 4. Save the product. ## Make a few shots Studio shot: ``` Keep the exact product shape and packaging from the reference. A clean hero shot on a pale background, soft side light, natural shadow. Keep all visible branding. ``` Lifestyle shot: ``` The same product on a warm wooden desk beside a window. Keep its proportions and label accurate. Leave clear space for a headline. ``` Check labels, small print, shape and reflections. Fix mistakes before upscaling. ## Make a product clip 1. Choose a still you like. 2. Describe one simple camera move. Keep the product still if its shape must stay exact. 3. Check the cost and click Create. 4. Watch the whole clip and check every part of the product. A spinning product made this way is generated video, not a 3D scan. For a real 3D model, see [3D models](https://videodraft.ai/docs/web-3d). ## Use the product Choose Use in Studio from the product’s menu to make images and video with it. - [Make video](https://videodraft.ai/docs/web-video-generation) - [Share your work](https://videodraft.ai/docs/web-sharing) --- # Make an avatar video > Pick a presenter, give it words and a voice, then render a video with captions. Source: https://videodraft.ai/docs/web-avatar-video ## Get ready Use a clear, front-facing portrait, or pick a saved avatar. Have your script ready, or a clean recording to use instead. The cost appears before you render. Short script: ``` Welcome to Northwind. Start with one clear brief, keep your references together, and review the first cut before making more. ``` ## Make the first version 1. Click Avatar Video in the sidebar. 2. Type the script under What should your avatar say?, or click Generate script with AI. To use a recording instead, click Upload audio. 3. Pick a presenter: Choose My Avatar, Create New Avatar or Use Original Image. 4. Pick a voice and click Preview voice to hear it. 5. Click the Generate avatar video button next to the script. Your avatar video opens on its own page. If you upload audio, listen to it first. Avoid clipped words, background noise and sudden starts. ## Check the performance - Listen for wrong words and pronunciation. - Watch the mouth and face through the whole clip. - Check pauses, framing and the last frame. - To change something, edit the script or voice, or use Change Avatar Look. ## Render with captions and download 1. To add captions, open Captions, turn on Enable captions and click Apply. 2. Click Render Video to make the final video. 3. From the export menu, choose Export with captions. 4. Download the file and watch it to the end. > **Presenter or scene?** > > Avatar Video is for someone talking to camera. For a character speaking inside a scene with movement and camera work, [make video](https://videodraft.ai/docs/web-video-generation) instead. - [Speech, music and sound](https://videodraft.ai/docs/web-sound) --- # Share your work > Share with a public link or invite people. Source: https://videodraft.ai/docs/web-sharing ## What to share | Work | How to share | | --- | --- | | An AI Studio project | Open its … menu and choose Share. | | AI Studio inside a video project | Click Open project to share, then share the video project with Share Project. | | Work in Character Studio or Product Studio | It is private to you. Download the results, or use them in a video project you share. | | A downloaded file | Send it like any file. Sharing settings do not apply to copies. | ## Share a view-only link 1. Open Share on the project. 2. Turn on Public link: anyone with the link can view. Only the owner sees this switch. 3. Copy the link and open it in a private browser window to see what others see. 4. Turn Public link off to stop sharing. People keep anything they already downloaded. ## Invite people 1. In Share, type an email address under Enter email address. 2. Pick Viewer to let them look, or Editor to let them work on it, and send the invite. 3. People with access and Pending invites show who is in. 4. The owner can change roles or remove people from the same list. > **Leaving and deleting** > > People who are not the owner see Leave session, which removes only their own access. When the owner deletes the project, it is gone for everyone. --- # Fix a problem in the web app > Find a missing result, fix a failed one, and avoid paying twice. Source: https://videodraft.ai/docs/web-troubleshooting ## A result failed or is still working 1. Open the AI Studio project where you made it. 2. If it is still working, wait. Clicking Create again starts, and pays for, a new job. 3. If it shows Failed, read the error. Check the required inputs, length, resolution and account. 4. Fix the cause, then click Retry. If an agent started the job, ask it to check the same job instead of starting another. ## A result is missing - Check you are signed in to the right account and workspace. - Open the right project. Results made inside a video project, character or product are saved there. - Turn off Favorites if it is on. - If only the download failed, download the finished result again instead of remaking it. ## A model will not accept the request Settings from the previous model may not work with the new one. Check the ratio, length, resolution and references after switching. If the model button says Requires image input, add an image first. If you use your own Fal or ElevenLabs key, see [using your own keys](https://videodraft.ai/docs/providers-and-usage#byok). ## A video project will not export Look for clips still generating, missing clips or scenes with errors. Finish or replace them, play the timeline, then export again. ## Contact support Include the project, when you made the request, the model and settings, and the exact error. Say whether the result exists and only the download failed. Never send API keys or passwords. - [Contact support](mailto:contact@videodraft.ai) - [How credits work](https://videodraft.ai/docs/providers-and-usage?mode=web) --- # MCP connector > Connect Claude, ChatGPT, Codex or Cursor to VideoDraft and make media without leaving the chat. Source: https://videodraft.ai/docs/mcp ## What the connector does The connector lets an AI app you already use work with your VideoDraft account. From the chat it can make images, video, voice, music and 3D models, run a video project, and read your projects and media. It uses the same account, credits and history as the [web app](https://videodraft.ai/docs/web-app). Images, video and sound it makes appear in AI Studio. It runs on VideoDraft servers, so there is nothing to install. The [CLI](https://videodraft.ai/docs/cli) is the terminal version of the same tools. | Setting | Value | | --- | --- | | Connector name | VideoDraft | | Server URL | https://app.videodraft.ai/api/mcp | | Transport | Streamable HTTP | | Sign-in | OAuth in your browser, or a personal access token | | Plans | Every plan, Starter included | ## Claude and Claude Code 1. In the Claude app, open Customize, then Connectors, then Add custom connector. 2. Enter the name VideoDraft and the server URL above. 3. Click Connect and finish signing in to VideoDraft in your browser. In Claude Code, add the server from your terminal, then run /mcp in a session to sign in. Claude Code: ``` claude mcp add --transport http videodraft https://app.videodraft.ai/api/mcp ``` ## Codex Add the server with this command, or in Codex’s MCP settings. Sign in when Codex asks. Terminal: ``` codex mcp add VideoDraft --url https://app.videodraft.ai/api/mcp ``` ## ChatGPT and Cursor ChatGPT: turn on Developer mode, then add a custom app with the server URL and OAuth. Your ChatGPT plan and workspace must allow custom apps. OpenAI’s guide, linked at the end of this page, has the current steps. Cursor: open the command palette, run Open MCP settings and add the entry below. Restart Cursor and sign in. ~/.cursor/mcp.json: ``` { "mcpServers": { "videodraft": { "url": "https://app.videodraft.ai/api/mcp" } } } ``` ## Use a token instead of a browser sign-in If an app cannot sign in through the browser, create a personal access token at app.videodraft.ai/mcp-keys and send it as a bearer header. Tokens start with vd_mcp_. Bearer header: ``` claude mcp add --transport http videodraft \ https://app.videodraft.ai/api/mcp \ --header "Authorization: Bearer vd_mcp_…" ``` > **You see a token once** > > VideoDraft keeps only a fingerprint of it, so it cannot show it again. If you lose it, create a new one. ## What your agent can reach Tools come in groups called lanes. Agents call get_tool_catalog with a lane name to get its tools and how to use them. | Lane | Covers | | --- | --- | | assets | Images, video, voice, music, sound effects, dialogue, 3D | | asset_io | Uploads, downloads and media descriptions | | projects | Create, read, update and version projects | | project_data | Scenes, shots, styles, visual assets, timeline clips | | production | Hosted production runs and video export | | library | Media, generation history, characters, products, styles | | account | Identity, credit balance, model lists, per-model costs | ## Credits and access - Generations spend the same credits as the web app and the Mac app. The connector itself costs nothing. - Ask for an estimate first. Video, batches of shot images, upscales and avatar renders cost the most. - Media made outside a video project goes into one AI Studio project for that connection. - Every call runs as you and can reach only your projects. - A sign-in lasts 30 days and renews in apps that support it. Revoke a sign-in or token at app.videodraft.ai/mcp-keys, where sign-ins appear as OAuth: . - If a request is rate limited, wait and try again. Do not resubmit the same job over and over. > **Detailed editing happens in VideoDraft Editor** > > The connector can make media, produce a video and place clips on a web project timeline. Precise cuts, color, captions and mixing happen in [VideoDraft Editor](https://videodraft.ai/docs/native-editor) on the Mac. - [MCP connector overview](https://videodraft.ai/mcp) - [Manage your keys](https://app.videodraft.ai/mcp-keys) - [Use the CLI instead](https://videodraft.ai/docs/cli) ## Try it Ask your agent: ``` Estimate one image of a blue ceramic cup on an oak desk. After I approve, make it, wait until it finishes, and tell me the AI Studio project and the image link. ``` Open that project in AI Studio to see and download the image. If the chat disconnects while the job runs, ask the agent to check the same job instead of starting a new one. - [AI Studio basics](https://videodraft.ai/docs/web-ai-studio) - [ChatGPT custom apps](https://platform.openai.com/docs/guides/developer-mode) - [Codex MCP setup](https://developers.openai.com/codex/mcp/) --- # Command line (CLI) > Install the videodraft CLI, sign in, and make video, images, audio and 3D models from your terminal or CI. Source: https://videodraft.ai/docs/cli ## Install and sign in The CLI is published on npm as videodraft. It needs Node 20.18.1 or newer. Install: ``` npm install -g videodraft videodraft --version ``` Sign in once in your browser. In CI, skip the login and set VIDEODRAFT_API_KEY to a token from app.videodraft.ai/mcp-keys. videodraft logout signs you out and clears the saved credentials. Sign in: ``` videodraft login videodraft whoami ``` ## Make something An image, clip, voiceover, music track or 3D model does not need a project. Add --estimate to see what a job costs before you run it. Examples: ``` videodraft generate image "a red fox in snow, cinematic" --download ./out/ videodraft generate video "slow dolly over a misty lake" --duration 6 --estimate videodraft generate voiceover "Welcome to VideoDraft" --download welcome.mp3 videodraft create "30s launch video for our espresso machine" --ar 9:16 ``` - Run videodraft generate --help for the rest: music, sound effects, dialogue, voice changer, dubbing, Seed Audio and 3D. - videodraft models lists what you can use. videodraft costs shows what a model costs. - Every command accepts --json. Exit codes are stable: 0 ok, 1 error, 2 usage, 3 sign-in, 4 not enough credits. ## Submit, check and download Image and video commands wait for the result by default. To manage jobs yourself, add --no-wait and keep the job ID. Check a job with status, or collect it with wait. If wait times out, the job keeps running on the server. Job control: ``` videodraft generate image "a blue ceramic cup on an oak desk" --num 1 --no-wait --json videodraft status JOB_ID --json videodraft wait JOB_ID --download "./media/{job_id}_{index}.{ext}" --json ``` > **Inside ADE** > > ADE already bundles the CLI and its agent skill. Agents there run the bundled copy using the exact path ADE gives each chat. ## Collect a small batch Estimate first. Submit each approved job with --no-wait and save its ID, then collect them all with one wait command instead of checking each job separately. Collect jobs: ``` videodraft wait JOB_ID_1 JOB_ID_2 JOB_ID_3 --download "./media/{job_id}_{index}.{ext}" --json ``` videodraft sessions current shows which AI Studio project your jobs go into. Use --session to pick a different one. A downloaded file is a copy; the original stays in your web history. ## Teach your coding agent to use it The CLI comes with an agent skill that covers model choice, references, waiting for jobs and costs. Install it and Claude Code, Codex and Cursor know how to make video without being told each time. The CLI finds which agents you have and installs the skill for each. Install the skill: ``` videodraft skills install ``` - [CLI overview](https://videodraft.ai/cli) - [Connect an agent over MCP](https://videodraft.ai/docs/mcp) --- # Workspace MCP > Built-in tools that let an agent in ADE read your other chats, tasks, diffs and terminal output. Source: https://videodraft.ai/docs/workspace-mcp ## What it is Every chat in ADE comes with built-in tools called videodraft_workspace. With them, an agent can see the rest of the app: other chats and tasks, what changed in the diff and what the Run tab printed. It can also leave review comments on the diff. - Every agent gets it, in VideoDraft mode and Code mode. There is nothing to turn on. - It only reads, except for review comments on the diff. - It stays on under Enterprise data privacy, because it runs on your Mac. > **Not the MCP connector** > > The [MCP connector](https://videodraft.ai/docs/mcp) lets apps outside ADE, such as Claude or ChatGPT, reach your VideoDraft account. Workspace MCP runs inside ADE and sees only ADE. ## Ask for it in plain words You do not need to name the tools. Say what you want and the agent picks them. It reads long chats a page at a time, so it takes in only what it needs. Examples: ``` Read what the design chat decided about the title card, then use it here. Search my past chats for the color grade we used on the Northwind promo. Review this task’s diff and leave inline comments on anything risky. The dev server in the Run tab crashed. Read its output and fix the cause. ``` --- # Guide for agents > Rules to give an AI agent that works with VideoDraft, plus the docs in machine-readable form. Source: https://videodraft.ai/docs/agent-guide ## Core rules - Work inside the assigned task. Save documents there, and images, video and audio under media/. - Load the videodraft skill before using VideoDraft tools. - Make single images, clips, voice, music and sound directly. They do not need a project. - Use VideoDraft Editor for timeline work and export when its tools are available. Use a web video project only when the user asks for one. - Check the live tool and model list instead of assuming names. - Estimate before large jobs, and check a running job instead of submitting it again. - Show finished media with a link to the local file, and put the key result first. ## Find the current tools Discovery commands: ``` videodraft --help videodraft tools list videodraft tools list --lane assets videodraft tools list --lane production ``` Inside ADE, use the bundled CLI path from the chat instructions in place of videodraft. Claude and Codex chats in ADE also get videodraft_editor tools, which control VideoDraft Editor even with no window open. ## Starter context Give this to an agent: ``` You are helping with a VideoDraft video-production task. Save documents in the task and generated or downloaded media under media/. Use the videodraft skill and check the live tool list before choosing tools. Make standalone assets directly without creating a project. Use VideoDraft Editor for timeline assembly and export when it is available. Estimate costs before large jobs. Show finished media with links to local files and put the key outcome first. ``` ## Machine-readable docs /llms.txt is a short index. /llms-full.txt is every page in one file; add ?mode=film, code, web or integrations for one path. Any page is available as Markdown by adding /llms.txt to its URL. --- # Make 3D models > Make a 3D model with an agent or the CLI and download it for your 3D app. Source: https://videodraft.ai/docs/web-3d ## Where 3D lives 3D models are not made in AI Studio. Ask an agent connected over [MCP](https://videodraft.ai/docs/mcp), use the [CLI](https://videodraft.ai/docs/cli), or ask an agent in ADE. A 3D model does not need a video project or an AI Studio project. You can start from a text description or reference images. Rigging, which adds a skeleton so the model can move, is a separate request. ## Make and download Ask your agent: ``` Check which AI models can make 3D. Propose one for a simple ceramic cup from my reference image and estimate the cost. After I approve, make it, wait until it finishes and download every mesh and texture file into one folder. List the files and formats. ``` Or from the terminal: ``` videodraft models 3d videodraft generate 3d --help ``` 1. Use a clear reference image or a short description of one object. 2. Check the AI model, output format and estimate. 3. Keep the job ID until the job finishes. 4. Download every file and open the result in your 3D app. ## Check the result - Turn it around and check the shape from every side. - Check that textures cover the whole surface. - Keep texture files next to the mesh when the format points to separate files.