Firehose for 10 December 2025
Speech recognition evaluations covered an index of ASR benchmarks, a transcription cleanup comparison across cloud audio models, and a test of Whisper fine-tunes against stock Whisper on local inference. The accompanying datasets were a small speech-to-text set of 92 audio samples with ground truth transcriptions and a fine-tune evaluation set.
GitHub Repositories
3 repos created
- Speech And ASR Evaluations GitHub · Evaluation Index repository for speech recognition and ASR evaluations
- Transcription Cleanup Eval 1225 GitHub · Evaluation Evaluating various cloud audio understanding models on the transcribe and cleanup workflow/task
- Whisper Fine Tune Accuracy Eval GitHub · Evaluation Comparing Whisper fine-tunes versus stock Whisper on local inference
Datasets
2 datasets created