Scribe v2 Realtime

Scribe v2 Realtime is built for live use where every millisecond counts. It ingests microphone or media streams and returns incremental hypotheses that settle quickly into punctuated, well-cased text. Word and segment timestamps stay consistent for editing and captions, speaker diarization can separate voices, and multilingual recognition handles code-switching without manual toggles. The API emits structured JSON so products can trigger actions, update subtitles, or store aligned transcripts in real time. Controls for endpointing, buffering, and confidence thresholds let you trade speed for stability, and lightweight models keep costs predictable for continuous sessions like support calls, meetings, and live broadcasts.