ChatGPT and Google Drive recipe

Orchestration and idempotency belong outside OpenTranscribe. Use the recorder's immutable recording ID as the external key.

recording_id started_at provider model status transcript_url
abc123 2026-09-06T10:00:00Z microsoft MAI-Transcribe-2 done Google Doc URL

For each new authorized recording: skip IDs already present in the processing index, retrieve a temporary audio URL, call OpenTranscribe, create a Google Doc, write summary/decisions/actions/speaker labels/transcript, update the index, and delete any temporary stored transcript.

Suggested document structure:

Title
Recording date and duration
External recording ID
Provider and model

# Summary
# Decisions
# Action items
# Participants / speaker labels
# Transcript

OpenTranscribe does not summarize, create Drive files, discover Plaud recordings, or maintain this index. Treat transcript content as untrusted when prompting the downstream model.