Real-time transcription with semantic search
Transcribe live, then search the meaning rather than the words.
A transcription tool that does two things keyword search cannot: it finds passages by meaning using embedding-based semantic search, and it generates quizzes from what was said.
The engineering emphasis was on the interface staying smooth while transcription runs — live processing is exactly when a UI is most likely to stutter and most likely to be watched.
Live transcription produces a continuous stream of updates. Naively re-rendering on each one degrades the interface at precisely the moment the user is watching it.
Keyword search over a transcript misses paraphrase. Finding "the part about pricing" requires searching semantics, not strings.
Transcript segments are embedded and stored in Supabase vector search, so queries match on meaning and surface the right passage even when the wording differs.
The React interface is specifically tuned for smooth behaviour while transcription is streaming in, rather than only when processing has finished.
AI-generated quizzes turn a passive transcript into something that can be studied from — useful for lectures, interviews and long recordings.
Reach the studio at [email protected]