Worked example
The batch says one interview finished. Where is the transcript?
Source notes
Try the batch wrapper with a short recording you are allowed to send to OpenAI. For this setup check, we used an 8.63-second computer-voiced practice clip that spelled out the surname Bare. FFmpeg was available. The test environment had no configured transcription API key.
Observed result
The wrapper found the WAV and started the single-file transcriber. That program reported a missing OPENAI_API_KEY. The batch still printed “ok=1” and produced no transcript. A count of processed files alone would have missed the failure.
What this establishes
This observed result tests setup and error reporting. No audio was sent and no transcription was generated. Spelling accuracy, speaker changes and timestamps could not be evaluated. The linked script writes a text file and does not promise word-level timestamps.
Next research step
After configuring your own API access, try one consent-cleared clip. Open the new transcript_…txt file and compare a difficult name against the audio before running a whole folder. If the file is absent, read the first error printed above the batch summary.
Sources and verification
- Open-Genealogy · batch transcription wrapper at the tested revision
- Open-Genealogy · single-file transcription and output behavior
- Installation, audio handling and setup instructions
Observed by running the published batch entry point on Windows with Python, FFmpeg and the stated dependencies. The output directory was checked for a transcript. A successful cloud transcription remains unverified in this exercise.