Lumiere
Prompt the cut. Watch the agent reason. Approve the timeline. Editing collapses from hours to minutes — every decision stays yours.
Every prompt produces a visible thought trace, tool list, and timeline diff. Bad assumptions die in the chat, not on your timeline.
Lumiere watches every clip and reads every transcript. Files get renamed, tagged, and grouped by scene, speaker, and emphasis — so the agent finds the right shot in milliseconds.
Built directly on WebCodecs. Open a tab, drop footage, work like it's a desktop NLE — and ship straight from the browser.
Captions, pacing, b-roll, motion graphics, color — the parts that usually eat your evening, handled in one browser tab and steered by a single prompt.
Auto-transcribed and word-synced, with karaoke highlight, clean wrapping, and brand styling — every line landing exactly when it's spoken.
Lumiere pulls the slack between words so the cut never drags.
Describe it — titles, lower-thirds, and callouts render live on the canvas, animated and on-brand.
The agent reads the transcript, watches your clips, and lays the right shot over the right line.
Log-aware color and real 3D LUTs, matched shot to shot so the whole video feels intentional.
WebCodecs renders locally — no upload, no render farm, no desktop install. Reframe once, export every aspect ratio.
Your timeline. Your taste.
AI handles the rest.
Every edit is visible.
Every decision is yours.
One plain line of intent — read like an assistant editor would read it.
The plan appears in chat before the timeline moves. Steer or veto.
Decisions land as real clips, real cuts, real captions — nothing locked.
Judgment stays where it belongs. The agent does the mechanical work.
Designed for creators who focus
on the vision, not the friction.
Private beta starts with capped free access. Pro and Max add monthly AI credits for heavier edits, transcription, and deeper analysis.
For testing the editor with daily AI limits and full timeline fundamentals.
For regular AI edits, captions, and assisted finishing with fewer beta limits.
For heavier motion, deeper analysis, and larger multi-step edit sessions.
Import: H.264 (MP4/MOV) and VP9/AV1 (WebM), up to 4K, 60fps — plus H.265 where your browser supports it. Audio: WAV, AAC, MP3. Export: MP4 (H.264; H.265 where supported), with bitrate control.
Yes. XML (FCPXML) and EDL export are in beta for Pro and Max tiers. Full timeline reconstruction in your desktop NLE.
No. Your media stays in your own private cloud storage bucket, and nothing you upload is ever used to train any model. The agent only reads your project to make the edits you ask for — never shared, never sold.
The editor runs in your browser, but both your media and the AI agent need a connection — the agent reasons in the cloud and your footage streams from your private storage. Your project autosaves continuously, so a dropped connection won't lose your edits.
Editors using Lumiere ship a finished talking-head video in 12 minutes instead of 3 hours.