Recorded Overview
Stable
Recorded vs. Realtime; when to use which.
Recorded Overview
Recorded is a Voice Context Layer mode for finished audio. Realtime is the other mode, for live speech.
Offline full-file transcription: submit an audio URL, process asynchronously, receive complete transcription results.
How it works
Submit audio URL (POST /v1/audio/transcriptions)
→ Probe audio format and duration
→ Slice into segments
→ Concurrent transcription (multiple chunks processed simultaneously)
→ Aggregate (merge into global timestamps)
→ Optional LLM review for error correction
→ Webhook callback (optional)
When to use
| Scenario | Suitable |
|---|---|
| Post-meeting processing | ✅ |
| Batch audio file transcription | ✅ |
| Need precise timestamps | ✅ |
| Need structured results | ✅ |
| Need subtitle files | ✅ |
| Need translation | Coming soon |
vs. Realtime
| Dimension | Recorded | Realtime |
|---|---|---|
| Transport | HTTP async | WebSocket streaming |
| Input | Audio URL | PCM16LE audio stream |
| Latency | Minutes | Milliseconds |
| Output | Complete JSON | Event stream |
| VAD | Not needed | Server-side |
| Best for | Post-processing | Live |
Current capabilities
| Capability | Status |
|---|---|
| Full-file transcription | ✅ |
| Segment timestamps | ✅ |
| Confidence scores | ✅ |
| Translation | Coming soon |
| Webhook callback | ✅ |
| LLM review correction | ✅ (optional) |
| Speaker diarization | Coming soon |
| SRT/VTT subtitles | ✅ (generated from timestamps) |
Capability boundaries
If a capability is not yet stable, this page marks it explicitly. Do not treat experimental outputs as stable.
Next steps
- Transcribe Audio — submit a job
- Timestamps & Speakers — timestamp details
- Subtitles — subtitle generation