Importing existing video
Bringing Loom, Zoom and file libraries into the pipeline.
Imported video runs through the same pipeline as native recordings, with one difference: it has no interaction metadata, so click moments are inferred rather than known.
Supported sources
- Direct connectors: Loom, Zoom cloud, Google Drive, Dropbox, Box, YouTube, Vimeo, S3
- File upload: MP4, MOV, WEBM, MKV, AVI up to 12 GB and 4 hours
- Audio only: M4A, WAV, MP3 — produces a written guide and an audio-led video
What inference can and cannot do
| Signal | Native recording | Imported video |
|---|---|---|
| Which element was clicked | Known exactly | Inferred from cursor dynamics |
| Element label for guide text | From accessibility tree | From OCR of the frame |
| View changes / chapters | From route events | From frame differencing |
| Typed input | Captured as intent | Not available |
| Zoom framing accuracy | Element bounds | Approximate region |
Inferred keyframes are flagged
The Studio marks every keyframe that came from inference so you can check them. In practice most are correct and a few need nudging.