Media Whisper

Context: $ARGUMENTS

Quick start

• Transcribe video → SRT: extract 16 kHz mono WAV (Step 2) → run ASR (Step 3) → SRT is written next to the input.
• Translate foreign speech → English SRT: Step 2, then whisper-cpp -tr ... or faster-whisper task="translate" (Step 3).
• Burn-in captions: generate SRT here, then hand off to the ffmpeg-subtitles skill.
• Soft-mux SRT into MP4: scripts/whisper.py srt-mux --video in.mp4 --srt out.srt --output tagged.mp4.

When to use

• Auto-generate captions (SRT/VTT) from audio or video.
• Translate non-English audio to an English transcript.
• Produce a searchable text transcript or word-level JSON for a podcast/interview.
• Build an ASR stage in a larger pipeline (precedes diarization, search indexing, summarization).
• Do NOT use for speaker identification — whisper doesn't do diarization reliably. Hand off to pyannote-audio (see references/whisper.md).