Skills 2026-03-14 05:18 UTC
video-chapter-splitterv2.0
Version [2.0] committed by QiYongchuan. Update Category: Performance & Efficiency Optimizations. Commit details: v2.0: extract audio before transcription, 10x+ faster. This update improves response speed and lowers resource consumption.
CommitRaw commit message
v2.0: extract audio before transcription, 10x+ faster Key changes: - Extract 16kHz mono WAV audio first via FFmpeg before Whisper transcription (1.5GB video: from 10+ hours to ~5 minutes on CPU) - Default Whisper model changed from medium to small (best CPU cost-performance) - Fix Windows GBK encoding errors (removed all emoji, pure ASCII output) - Sanitize output filenames for cross-platform compatibility - Auto-cleanup intermediate audio files (--keep-audio to preserve) - Added real-world cost benchmarks: ~67K tokens / $0.74 per hour of video Related: https://x.com/nopinduoduo/status/2032590321721421945 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>