Cloud recording on our Video SDK account works reliably (MP4 is always produced). The audio transcript (VTT) is produced for some sessions and silently not produced for others. When it’s absent, there is no TRANSCRIPT file, no audio_only file, and the recording.transcript_completed webhook never fires, so there is nothing for us to ingest.
We were recently told by support that the “Audio transcript” setting does not appear in our Video SDK account. That is contradicted by our own data: Zoom has generated and served audio-transcript VTT files for this account, most recently on 2026-06-28, downloadable from zoom. So audio transcription is clearly provisioned on the account. The problem is that it fires inconsistently, and we cannot find any documented condition that explains when it does or doesn’t.
EVIDENCE TRANSCRIPTION IS ACTIVE ON THIS ACCOUNT
Multiple Video SDK sessions received a Zoom-generated TRANSCRIPT VTT via recording.transcript_completed, e.g.:
- Session ended 2026-06-26 13:55 UTC, ~9 min, transcript delivered, served from zoom.
- Session ended 2026-06-28 17:57 UTC, transcript delivered, served from zoom.
FAILING EXAMPLE (no transcript ever produced)
- session_id: MuGn5YQpQf6U+jH2lvL68Q==
- session_name: 019f13f9-2467-78e3-a0fd-90fa6e0a2b7a
- Recorded 2026-06-29 15:57-19:15 UTC, ~198 minutes, 524 MB.
GET /v2/videosdk/sessions/MuGn5YQpQf6U%2BjH2lvL68Q%3D%3D/recordings returns only:
- MP4 (shared_screen_with_speaker_view) — status completed
- CHAT (txt) — status completed
No TRANSCRIPT file, no audio_only file. end_time is null on the session object. More than [N] hours after the MP4 completed, no transcript has appeared and no recording.transcript_completed event has fired.
OBSERVED PATTERN
Across our recorded Video SDK sessions, the ones that received a transcript were all short and ended cleanly. The ones that received no transcript are the long sessions (this 198-minute one, and another ~312-minute one). We cannot find this duration/behavior limit documented anywhere.
WHAT WE NEED ANSWERED
-
Audio transcription is generating VTT files for this account (proof above). Why is it not generated for the recording listed, and what is the actual condition that gates it? (Recording duration cap? Requirement for an audio_only file? Recording layout? Force-ended / unfinalized sessions?)
-
If there is an account or app setting governing this, tell us exactly where it lives for a Video SDK account, since it does not appear in our settings UI where regular Zoom Meetings exposes “Create audio transcript.”
-
For the specific failed recording (MuGn5YQpQf6U+jH2lvL68Q==), can a transcript be generated retroactively, or is it permanently unavailable?
-
Confirm whether recording.transcript_completed is the only delivery mechanism for the VTT, or whether it can also be pulled on demand from the recordings API once generated.
We have been chasing reliable transcripts on this account for an extended period. We need a definitive answer on the gating condition, not general recording-settings troubleshooting.