Get transcript
00:01
00:05
00:09
00:13
00:17
00:21
Recorded and live streams. Verbatim transcription in the spoken language. Find key moments, cut 16:9 or 9:16 clips with burned captions, all under a hard spend cap.
Runs on your own infra · captions-first so most transcripts are free · per-agent quotas & a daily budget cap
Built for teams publishing volumes of video — cheap by default, accurate when it matters.
Serve YouTube captions and a permanent cache first — most requests cost nothing. Fall back to AI only when captions are missing.
One shared pipeline per stream. Verbatim text as it airs, music/silence skipped, resilient to reconnects.
Set in/out, nudge by ±0.5s, build a frame, render to MP4 — 16:9 or 9:16, with styled burned-in captions.
Automatic chapters from pauses and transitions. Click a moment to jump the video.
Full-text search across every stored transcript; jump straight to the timestamp.
Email/password accounts, admin approval, per-agent quotas, and per-user usage.
Daily $ cap and quotas stop runaway cost. VAD + dedupe + batch keep it low.
Download transcripts as TXT / SRT / VTT and clips as MP4 — or archive to your storage.
Deploy the web app and the render worker yourself — your data stays yours.
Paste a link. Get text and clips. No fragile setup.
Watch, youtu.be, /live/ or Shorts. We detect live vs recorded and check the cache.
Captions appear instantly (free). If none exist, transcribe with AI — you approve the cost first.
Mark a range (or a whole frame), render to MP4 with captions, download or archive.
You run it; you only pay your infra + model usage. Typical team lands at a few dollars a month.