The speech recognition service is missing important improvements to the TigerFlow-ML transcription task. In particular, the transcription task has significantly improved how windowing is handled for files exceeding the built-in 30s Whisper chunk length. We need to update the service API to incorporate this and other updates so that they produce the same output.
The speech recognition service is missing important improvements to the TigerFlow-ML transcription task. In particular, the transcription task has significantly improved how windowing is handled for files exceeding the built-in 30s Whisper chunk length. We need to update the service API to incorporate this and other updates so that they produce the same output.