Commit Graph

2 Commits

Author SHA1 Message Date
nnnet 3ea11b4e5b whisper.youtube: sliding-window streaming + live mic mode for Parakeet
- parakeet_transcribe.py: rewrite --stream as sliding window with overlap
  dedup (case-/punctuation-insensitive); add --window and --shift params
- parakeet_mic.py: live mic capture via pw-record/arecord, progressive
  transcription with same sliding-window algorithm
- parakeet: export NUMBA_CACHE_DIR to work around librosa caching error
  on the mounted filesystem

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-11 21:36:01 +03:00
nnnet 7a4ff7ec36 add examples/whisper.youtube: YouTube → speech-to-text pipeline
- yt_transcribe.py: yt-dlp + ffmpeg + whisper-cli (large-v3) wrapper
- parakeet_transcribe.py: alternative NVIDIA Parakeet ASR backend
- parakeet: shell launcher for parakeet via local venv
- transcriptions/q1CluXct_VI.txt: sample output (8h audiobook)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-11 20:32:07 +03:00