Turn rough footage
into finished video.
A free, local editor for you. An API and MCP for your agents. You both work on the same project.
Edit it yourself. Or ask your agent.
It is the same project either way. A cut you make by hand and a cut your agent writes land in the same edit, so neither of you starts over.
Cut the words, not the timeline.
The transcript is the edit surface. Select a sentence, press delete, and the cut is made. Free on macOS and Windows, and local footage never leaves your machine.
The same edit, as data.
Upload, compose, render: three calls over REST, MCP or the CLI. Change one line and only that one thing changes. One command wires every agent on your machine.
npx -y cueframe install{
"type": "clip.add",
"clip": {
"id": "l1",
"startTime": 15,
"duration": 5,
"source": {
"kind": "overlay",
"primitiveId": "lower-third",
"params": {
"text": "Daniel Marsh",
"subtitle": "Host",
"tokens": { "accentColor": "#f7c948" }
}
}
}
}Wires into Claude Code, Cursor, Codex, VS Code and 10 more.
Claude Code
Cursor
Windsurf
Zed
Gemini CLI
GitHub Copilot CLI
Cline VSCode Extension
OpenCodeAsk for a change. Watch it land.
A plain-English request, the timeline patching itself, and the Director checking its own work against real rendered frames. You could have grabbed the playhead at any point.
What else it does.
Every clip on this page is a screen recording of the shipped app.
Word-synced captions
Captions come from the transcript's word timings, so every word lands exactly when it's spoken.
Motion titles and lower thirds
Cinematic titles and overlays from the same catalog the renderer uses, right in the script.
Export in any format
MP4 up to 4K, SRT and VTT, or EDL and FCPXML for Premiere, Final Cut and Resolve. Local exports are free.
Nothing leaves your machine until you ask.
In the desktop app, importing, transcribing, editing and exporting all run locally. Your footage is not uploaded, and you can cut a whole video without sending us a frame of it.
You send a project up only when you choose to: a cloud render, or a project you hand to your agent. That choice is the line, and you draw it.
Most video APIs fill in a template. CueFrame reads what it renders.
When you do send a project up, the render earns the trip. The models read the footage you sent, and that reading is where these come from.
The camera follows whoever's talking
Speakers are found and tracked, so the frame moves with the conversation instead of cropping it.
Captions go behind the speaker
Composited from the footage itself. No green screen, no masking pass, no cutout look.
Every word has a timestamp
Word-level timing and speaker labels, so an edit can target a sentence.
Every aspect ratio from one edit
The reframe is content-aware, not a center crop. Never re-edited per platform.
Bring footage. Or start from a brief.
Every frame is a real render. Every shot came from licensed stock. Click a card to read the prompt behind it.
The editor is free. Cloud renders are $0.50 a minute.
Editing, transcription and exports run on your machine at no charge, and cloud extras like stock search and edit validation are free too. Credits are only spent when CueFrame's cloud does heavy work: rendering by the output minute (a three-minute cut is $1.50, 4K counts double, 720p half) and AI generation per asset. No subscription, no seats, prepaid credits that don't expire, and $20.00 free to start, no card.
Questions, answered.
Do I need to know video editing?
No. You describe the video you want; your agent writes the edit and CueFrame does the editor's work: framing, captions, pacing, brand. If you can review a draft, you can ship video.
What footage does it work with?
Real footage: a talk, a podcast recording, a product walkthrough, a founder update. Hand the API a file or a URL and CueFrame transcribes it and finds the speakers so the edit can follow them. Or drop footage straight into the desktop app, where it transcribes on your machine and stays there.
Is my footage private?
The desktop app is local-first: footage, transcripts and edits stay on your device and are never uploaded unless you put a project in the cloud. CueFrame uses cloud processing only where local can't do the job properly, like the state-of-the-art models that track speakers for a render, and only on what you send it.
Which AI tools does this work with?
Run npx -y cueframe install and it wires CueFrame into every agent it finds on your machine: Claude Code, Cursor, Codex, VS Code, Windsurf, Zed, Gemini CLI, Copilot CLI and more. Claude Desktop and ChatGPT can't be wired from the command line (neither reads a local MCP config), so add api.cueframe.ai/v1/mcp as a custom connector instead. There's a plain REST API too.
What does the desktop app add?
A native editor for macOS and Windows. Cut from the transcript, scrub the timeline, and ask the built-in AI Director for changes. It opens the projects your agent composes, and anything you fix by hand stays in the same edit your agent reads.
What does a video cost?
The desktop app is free: editing, transcription and exports all run on your machine at no charge, and cloud features like stock search and edit validation are free too. You pay when the cloud produces the video. Rendering bills by output minute, rounded up: anything up to a minute costs $0.50, a three-minute cut $1.50, 4K counts double, 720p half. You start with $20.00 in free credits, no card.
Will the same edit always produce the same video?
Yes. The edit is data, and rendering it is repeatable: pin the edit and you get the same video back. Writing the edit is the creative part, so an agent may draft it differently run to run; but once it's written, nothing re-rolls. Change one line and only that one thing changes.
Can I open the edit in Premiere or Final Cut?
Yes. Export the timeline any time and finish by hand. No lock-in.