A video studio you direct in plain English
MetaRush is an AI-powered video creativity and editing harness built on video as code. Drop in raw footage — takes, interviews, B-roll, screen recordings — and ask for “an intimate 60-second story, vertical, with tasteful captions.” AI identifies the scenes, faces and speakers, Claude Code assembles the cut from reusable components, and you get a grounded, explainable, reproducible video rendered to a real MP4.
In private preview. Runs on your own machines: macOS and Linux.
Video as code
In MetaRush, a cut isn’t a pile of clicks on a timeline. It’s a precise, versioned description of the video — every shot, line, caption, graphic and sound cue — rendered from components. That makes every video reproducible, every change reviewable, and every version recoverable.
Edits work like pull requests: each change is previewed first, confirmed by a person, and saved as a new immutable revision. Work on alternate versions side by side, lock the parts you love, and hand the finished cut to DaVinci Resolve, Final Cut Pro or Premiere when you want to finish by hand.
AI that understands your footage
Your choice of speech model
Every clip is transcribed on your own computer with word-level timing and speaker turns, using the speech-to-text model you choose. NVIDIA Parakeet and Whisper are supported today, and you can switch models as quality improves. Nothing leaves your machine.
Speaker identity
Voices are recognized across every video in a project. When it meets a new voice, MetaRush asks who it is instead of guessing — and then follows directions like “use only the guest’s answers.”
Scenes, shots and faces
Scene changes, shot types, on-screen text and who is on screen are indexed automatically, and faces are recognized across clips. Vertical and square cuts keep the speaker’s face in frame, and contact sheets show the story at a glance.
Asks instead of guessing
Unknown voices, music without known rights, a project with no goal or audience — MetaRush raises them as to-dos instead of quietly making a choice for you.
Motion graphics, inside the edit
Motion graphics aren’t a separate tool or a hand-off. Lower thirds, kinetic quotes, data cards with sources, end cards and word-emphasis captions come in your brand’s colors, and they preview in the same timeline as the cut.
When no template fits — a route drawn across a map, a counter that ticks up, a chart that builds — Claude Code writes a custom motion graphic for your project. Each one is checked so it can only use what it’s allowed to, previewed in context, and rendered only after a person approves that exact version.
Built on Remotion
MetaRush renders through Remotion, so every caption, title, layout and graphic is a real component rather than a baked-in effect. That gives you frame-accurate rendering, a preview that matches the final video exactly, and a library that grows: Claude Code can write new components for a project, and a team can build up its own templates in its own style.
The whole creative process
Story editing
Describe the story you want and get a cut that explains itself: why each line is there, where each moment comes from, and how long it runs. Reorder, trim or swap lines in plain English.
Narration and voiceover
Write or import a voiceover script and place each line where it belongs, then record your own or generate an AI voice (ElevenLabs), with the cost shown first and unchanged lines never paid for twice.
Sound and music
Camera sound, narration, music and effects on their own lanes. Turn each person up or down only while they speak, duck music under dialogue, and let music fit the video by itself. Multi-camera and multi-mic recordings sync automatically.
Conversations and podcasts
Bring in Zoom, Riverside or SquadCast recordings: each person’s audio lines up and gets its own level and cleanup, the automatic cut follows the conversation in five styles, and MetaRush finds the best moments for shorts.
Quality checks before delivery
Every render is decoded and checked against its plan: loudness and peaks against platform conventions, captions and overlays, and silence where there should be speech. Fixes are proposed, never applied silently.
Brand kit and resources
Keep logos, colors, fonts and slide templates in one place, and share footage, graphics and approved cuts across projects without copying anything.
One studio across your machines
A local dashboard on every computer shows the same projects, revisions and renders. Media lives in your own cloud storage, and heavy jobs wait politely while a laptop is on battery.
Coming next
- Generated shots — fill a missing shot with an AI video model (Google’s Gemini first, Runway next), in draft quality while you iterate and final quality for delivery, within budgets people set.
- Looks, with consent — generated shots that keep a person recognizable from a few stills and clips, only with their consent, and always disclosed.
- Recipes — repeatable workflows, like an interview episode from import to cleanup to cut to shorts, run with one request.
- Zoom cloud import — pull recordings straight from Zoom, with each participant’s audio.
- Object tracking for B-roll — follow the subject of a moment, not just faces.
- Visual understanding everywhere — the on-screen index on Linux and Windows, not just the Mac.
- More one-click actions — cut, cleanup, sound and shorts as action cards in the dashboard’s Claude Code panel.
Directed through Claude Code
MetaRush is skills-first. A set of Claude Code skills — ingest, footage review, story editing, revision, motion design, voiceover, QA, render and delivery — drive a single library of editing operations that the command line, the dashboard and an MCP server all share. Anything you can do in the dashboard, an agent can do too, and nothing changes without your confirmation.
Get early access
MetaRush is in private preview. If you make video — and want an AI collaborator that understands your footage and never edits behind your back — get in touch.
Contact JonCopyright 2026 Metavert LLC