Two years of Looms, rebuilt in an afternoon
You already have the recordings. They are scattered across Loom, Zoom cloud, Google Drive and someone's desktop, they are twenty minutes long, they have the wrong logo, and nobody can find them. Import them and run the same pipeline over the lot.
Drop video files, or paste a Loom / Zoom / Drive link
MP4 · MOV · WEBM · MKV · up to 12 GB
- 12 GB
- Max file size
- 8 sources
- Direct connectors
- Bulk
- Import a whole library at once
- Same
- Pipeline as native recordings
Re-cut, don't re-record
An eighteen-minute all-hands segment becomes a three-minute feature video, with the rambling removed and the branding applied.
Bring the whole library
Bulk import from Loom, Zoom, Drive or a folder of files. Everything lands in one searchable library with transcripts.
Suddenly findable
Every imported video is transcribed and semantically indexed, so “the part where we explain rate shopping” actually finds it.
Paste a link or drop a folder
Connect Loom, Zoom cloud recordings, Google Drive, Dropbox, Box, YouTube, Vimeo or an S3 bucket and pull an entire library across in one operation. Or drag a folder of MP4s onto the page. Transcription, chaptering and indexing happen on ingest, so the library is useful before you have edited anything.
- Direct connectors for Loom, Zoom, Drive, Dropbox, Box, YouTube, Vimeo, S3
- Bulk import with folder structure preserved as collections
- MP4, MOV, WEBM, MKV, AVI up to 12 GB per file
- Resumable uploads that survive a closed laptop
Drop video files, or paste a Loom / Zoom / Drive link
MP4 · MOV · WEBM · MKV · up to 12 GB
Editing decisions from footage that has no metadata
Native recordings carry interaction data. Imported footage does not — so the pipeline infers it. Cursor trajectory, dwell and acceleration identify click moments. Frame differencing identifies view changes. OCR reads on-screen labels so the guide can name the button you pressed. It is not quite as exact as native capture, and the studio shows you which keyframes were inferred so you can check them.
- Click inference from cursor dynamics
- View-change detection via frame differencing
- On-screen text recognition for step naming
- Inferred keyframes flagged distinctly for review
Re-brand a hundred videos at once
Select a collection, apply a brand kit, and queue. Every video is re-cut with the new intro, outro, lower thirds, cursor style and music bed, re-rendered, and republished at the same URLs so nothing you have already embedded breaks. Useful after a rebrand, a pricing change, or the day you finally stop using the old logo.
- Apply a brand kit across an entire collection
- Republish in place — existing embeds keep working
- Bulk translate a collection into new markets
- Queue runs off-peak with email notification on completion
The technical detail
- Formats
- MP4, MOV, WEBM, MKV, AVI, plus M4A / WAV / MP3 for audio-only
- Max file
- 12 GB per file, 4 hours duration
- Connectors
- Loom, Zoom, Google Drive, Dropbox, Box, YouTube, Vimeo, S3
- Bulk limit
- 500 files per import operation
- Transcription
- Automatic on ingest, 34 languages, speaker-separated
The written guides are the sleeper feature. Every video we publish also becomes a help article, in six languages, with screenshots that are always current. Support ticket volume on onboarding fell by half.
About import & re-cut
The polish is the same — branding, voice, captions, pacing. What differs is zoom precision: native capture knows which element you clicked, while imported footage infers it from cursor behaviour. In practice that means occasionally nudging a keyframe. Resolution is also capped by the source; we can upscale, but we cannot invent detail that was never recorded.
Yes, and this is a common path. Silent screen captures get a script written from the interaction sequence, which you then edit, and a studio voice reads it. Teams use this to turn QA recordings and bug reproductions into proper walkthroughs.
Pairs well with
In production at