Use cases
AI review pipelines
Get every call back as a transcript by speaker, or send each participant's audio, on its own track, to your own summarisation or evaluation models.
The problem
AI features that work on calls need the media afterwards, cleanly separated by speaker. Recording products often hand over one mixed file, and only after processing.
Why it fits
The recording already exists, separated by track, the moment the call ends.
How Synento helps
- Each participant's microphone is its own track, read back through the playback API, so there's no diarisation guesswork.
- The media is readable as soon as the call ends. Start from the
session.endedwebhook. - Your coding agent can wire up the integration and check from the recording that each participant's audio really arrived.
The architecture with Synento
Browsers publish camera and microphone over one WebSocket. The relay appends each participant's tracks to durable streams as they pass. Your pipeline starts from the session.ended webhook, reads each mic track through the playback API, and feeds per-speaker audio to your own models, or turns on Synento's post-call transcripts and reads them on the transcript.ready webhook.
Verify via the session manifest
After a two-participant test call, read the session manifest and confirm both participants have a cam and a mic track with a non-null start_ts. A track appears only once media reached it, so a missing track is a failed integration, not a quiet success.
Check the recording
# Did media really arrive? Ask the recording.
curl -s https://api.synento.com/v1/sessions/$SESSION_ID/manifest \
-H "Authorization: Bearer $SYNENTO_API_KEY" \
| jq '.data.participants[] | {user_id, tracks: [.tracks[] | {kind, start_ts}]}'Record the calls your product runs on.
Sign up for a free account. Replay, MP4 export and post-call transcripts are included, and a coding agent can set the integration up from llms.txt.