Faces, on-screen text, speech
Video intelligence on the timeline
Recognition, burnt-in text and speech arrive as tracks with timecodes. Click a name, a caption or a line and the player is already there.
- Metadata
- AI Segmentation
- Chapters
- Transcription
- Faces7
- Text Recognition
- Clips
Text recognition
c
3 occurrences
SortBy timeBy appearances
- CICICCARELLI1×
- SESELECTION1×
- BABackcountry1×
Hover a bar for detail · click to jump
Transcription
All speakers
Filter transcript…
00:12:04Speaker 1
We came into the second half knowing the next ten minutes decided it.
00:12:19Speaker 2
And that is the pass everyone will be talking about tomorrow.
00:12:31Speaker 1
He saw it early. That is the whole difference.
Nothing on this timeline was typed by anyone.
Recognition runs inside the ordinary ingest workflow, as a task like transcoding or proxy generation. You are notified when a job lands, and a rejected analysis fails fast with the real cause recorded instead of retrying in silence.
Four things the machine reads
Faces
GPU-accelerated recognition lists each person detected, with a name where the face is known, and a timeline of every appearance. When several recognised people share the screen, the track splits into one lane each.
On-screen text
Lower thirds, chyrons, slates, signage and captions burnt into picture, detected frame by frame with timecodes. Search the text, then seek straight to the frame it appears on. Video assets only.
Speech
A filterable transcript with timecodes and speakers separated, synchronised to the player. It is a Studio track too, so a quote can be cut without scrubbing for it.
Summary and entities
A written summary of what happens in the asset, the entities named in it with how often each occurs, and the detected spoken language with a confidence figure.
Bring one interview or one panel recording. We will run it and read the tracks with you.
Talk To Us