I use Omi with Custom STT pointing to my own local Whisper server, but it only handles live transcription. When I open a saved recording and tap Process Now, or when Omi syncs recordings saved while offline, Omi appears to send the audio to Omi’s backend for transcription instead of using my Custom STT endpoint.
Please let Custom STT process saved recordings too, including:
Process Now on a saved recording
Transcribe Later recordings
Recordings uploaded during offline/backlog sync
Ideally, these recordings should follow the same Custom STT configuration as live audio, with a clear status/error indicator and no upload to Omi’s transcription service when Custom STT is selected. A local-only option would be especially useful for users who want to keep audio on their own network.
This seems related to, but distinct from, the live-audio routing work in Omi issue #10431 (https://github.com/BasedHardware/omi/issues/10431) and PR #10447 (https://github.com/BasedHardware/omi/pull/10447). There is also issue #8006 (https://github.com/BasedHardware/omi/issues/8006) concerning offline recording sync and transcription. I opened issue #20174 (https://github.com/BasedHardware/omi/issues/20174) for this saved-recording/custom-STT gap.
Expected behavior: Saved and backlogged audio is transcribed through the configured Custom STT endpoint, just like live audio, rather than being routed to Omi’s transcription backend.
