OpenEdit is a free, agent-driven video editing pipeline that runs on your own machine. There's no GUI and no timelineL you describe the edit to your coding agent, and OpenEdit composes it in HTML and CSS and renders it to video using VEED Engine. This guide covers what OpenEdit does, how to install it, and what to expect while it's in beta. It's built for developers, creators, and agencies who already work with AI agents and want video production inside that workflow.
What Is OpenEdit?
OpenEdit is two things: an editing pipeline your AI agent drives, and the HTML-to-video renderer it renders with.
You run one install command, open your agent, and describe the job. The pipeline handles composition in HTML and CSS, and VEED Engine turns that composition into video locally on your machine — with no headless browser in the render path.
Because composition doesn't depend on the renderer, you aren't locked to it. If running a closed binary is a dealbreaker, you can render through Chrome instead.
OpenEdit is available now on GitHub
What Can OpenEdit Do?
Captions are the most polished path today, but they're one job out of several. Everything below runs through the same prompt-and-render loop.
Capability | What that means |
Edit, cut and reframe footage | Trim, restructure and reframe clips for a different aspect ratio, described in a prompt rather than dragged on a timeline |
Layer motion graphics and visual elementsTurn slides or websites into video | Charts on the numbers, titles, lower thirds, overlays and transitions, composed in HTML and CSS |
Turn slides or websites into video | Hand it a deck or a URL and get a video out — no screen recording, no manual timing |
Capture web pages | Pull a live page into a composition as footage |
Add captions and subtitles | Transcribe the audio, design a caption style that fits the footage, and burn it in |
Pull in generation services and MCP servers | Point OpenEdit at any video or image generating service, or any MCP server, and it brings the result into the composition |
Source video is optional. Stills, slides, generated media, or pure motion graphics are enough on their own when the brief calls for it.
Note: OpenEdit edits, it does not generate. If you want a talking head from a script, that's generation rather than editing — the VEED MCP handles that with Fabric, which uses credits like most other text-to-video models.
Who Can Use This Feature?
Price: Free to use. No subscription, no paywalls
Operating system: Apple Silicon Mac on macOS 26 (Tahoe) only — not Intel, not earlier versions of macOS
Agent: Agent-agnostic. Tested with Claude Code, Codex, and Gemini
Account: One free veed.io account, used for transcription only
Windows and Linux support is something VEED intends to build. How it gets prioritised depends on demand.
How to Set Up and Use OpenEdit
Install the skill
Run
npx skills add veedstudio/open-editin your terminalThis is the only setup command you need — your agent handles the rest
Open your agent
Launch Claude Code, Codex, Gemini, or whichever agent you use
The skill prepares the runtime and loads its
AGENTS.mdexplicitly, so it isn't tied to one model
Describe the job
Prompt your agent with what you want, for example:
Add subtitles to my video [VIDEO]Or run the skill directly on a selected video with
/open-editYour agent works out what to do, composes it, and renders it
Review and keep going
One to three minutes later you get a finished MP4 back, and a preview window opens automatically
Keep prompting to refine the result
Example follow-up prompts
Make the yellow darker, move the text up a bitAdd motion graphics to the revenue numbersReframe this to 9x16 and keep the speaker centeredTurn this deck into a video with titles between each slideUsing the Figma MCP, take this layout and build a branded subtitle style for my podcastRun 6 agents in parallel, each with its own caption system
What You Can Build with OpenEdit
Daily social output. Ship a week of Reels, TikToks, Shorts, or podcast cuts straight from your agent
Branded assets that stay on brand. Give OpenEdit a reference image or your brandbook once, and every clip after it matches
A client's week, done in an afternoon. Run the batch through a multi-agent session
Video inside the tools you already run. Connect any MCP or API and make OpenEdit part of your existing workflow
How OpenEdit Compares
VEED built its own HTML-to-video renderer and runs it locally on your machine. You style compositions in HTML and CSS, but there is no headless browser in the pipeline. Most agent-native video tools drive Chrome instead.
Metric | OpenEdit | Comparable browser-rendered workflows |
Render speed | Up to 2.2x faster | Baseline |
Pipeline | HTML-to-video renderer, no headless browser | Headless Chrome in the render path |
Where it runs | Locally, on your machine | Varies, often cloud or local browser |
Source: VEED's own early internal benchmarks, not a published comparison. Reproducible benchmarks and the analysis behind them are coming in the next few weeks.
Is OpenEdit Open Source?
Yes — the OpenEdit pipeline is open source under Apache-2.0. You can add your own APIs and MCPs, mix it into an existing skillset, or connect it to a pipeline you already run.
Only VEED's renderer binaries are closed source, under PolyForm Shield 1.0.0, and they're free to use: you can sell what you make with them at no cost to VEED. You aren't stuck with them either, since composition doesn't depend on the renderer.
Both licenses are in the repo.
Important Notes
OpenEdit is in beta. Here's where it stands today:
Apple Silicon Macs on macOS 26 (Tahoe) only. Not Intel, not earlier macOS
Captions are the most polished path. Reframing, motion graphics, slides-to-video, and web capture all work, and we're working on making them as flawless as captioning. Worth pushing on regardless — that's where feedback is most useful right now
It edits, it does not generate. OpenEdit composes and renders. Point it at a generating service or MCP server and it will pull the result in, but it won't make footage for you
Transcription uses your VEED account. Your audio — not the video — is uploaded and stored in order to transcribe it. Your limits are whatever your VEED account already has
FAQs
Q: Is OpenEdit free?
Yes. There's no subscription and no paywall. You only need one free veed.io account for transcription.
Q: Do I need a VEED account to use OpenEdit?
Only for transcription. The default captioning workflow uses VEED's transcription, which requires an account. Other jobs don't.
Q: Does OpenEdit work on Windows or Linux?
Not currently. OpenEdit runs on Apple Silicon Macs on macOS 26 (Tahoe). Windows and Linux support is something VEED intends to build, prioritised by demand.
Q: Which AI agents does OpenEdit work with?
It's agent-agnostic and has been tested with Claude Code, Codex, and Gemini. The skill loads its
AGENTS.mdexplicitly, so it isn't tied to one model.
Q: Can OpenEdit generate video from a script?
No. OpenEdit edits and renders. For a talking head from a script, use the VEED MCP with Fabric, which uses credits.
Q: Do I have to use VEED's renderer?
No. Composition doesn't depend on the renderer, so you can render through Chrome instead if a closed binary is a dealbreaker.
Q: How long does a render take?
Typically one to three minutes from prompt to finished MP4, depending on the job.
Q: Is my video uploaded to VEED?
No. Only your audio is uploaded, and only for transcription. Everything else renders locally on your machine.
Tips for Best Results
Be specific in your first prompt. Aspect ratio, tone, and brand references up front save a round of edits
Give it your brandbook once. A reference image or brand guide means every clip after it matches
Keep prompting after the first render. The preview window opens automatically — iterate from there rather than starting over
Push on the newer paths. Reframing, motion graphics, slides-to-video and web capture are functional but less polished than captions, and feedback on them is especially useful right now