Mathieu Virbel 7030e0f236 fix: optimize parakeet transcription batching algorithm (#577) ...

* refactor: optimize transcription batching to accumulate speech segments

- Changed VAD segment generator to return full audio array instead of segments
- Removed segment filtering step
- Modified batch_segments to accumulate maximum speech including silence
- Transcribe larger continuous chunks instead of individual speech segments

* fix: correct transcribe_batch call to use list and fix batch unpacking

* fix: simplify

* fix: remove unused variables

* fix: add typing

2025-08-27 10:32:04 -06:00

..

modal_deployments

fix: optimize parakeet transcription batching algorithm (#577)

2025-08-27 10:32:04 -06:00