wavecap-hallucination
Configure WaveCap hallucination detection and prevention. Use when Whisper outputs gibberish, repeated phrases, or phantom text on silent audio.
npx skills add majiayu000/claude-skill-registry --skill wavecap-hallucination-tobiaswooldridge-wavecap --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
# WaveCap Hallucination Prevention Skill Use this skill to detect and filter out Whisper hallucinations - false transcriptions generated when there's no actual speech. ## What Are Hallucinations? Whisper sometimes generates plausible-sounding text when fed: - Silent audio or white noise - Background music or tones - Very low-quality speech Common hallucination patterns: - "Thank you for watching" - "Transcription by CastingWords" - Repeated phrases like "the the the the" - Random punctuation or symbols ## Configuration Location Hallucination settings are in the `whisper:` section: - **User config:** `/Users/thw/Projects/WaveCap/state/config.yaml` ## Silence Hallucination Phrases Phrases that indicate Whisper hallucinated rather than transcribed: ```yaml whisper: silenceHallucinationPhrases: - "thank you" - "thanks for watching" - "transcription by castingwords" - "casting words" - "all right here we go" - "alright here we go" - "all right let's go" - "alright let's go" - "you" - "bye" - "uh" - "hmm" ``` When these phrases appear on low-energy audio, the transcription is discarded. ### Add Custom Hallucination Phrases If you notice recurring false phrases, add them: ```bash # View c
- What Are Hallucinations?
- Configuration Location
- Silence Hallucination Phrases
- Add Custom Hallucination Phrases
- Segment Repetition Detection
- Blank Audio Detection
- View Current Settings
- Diagnose Hallucination Issues
- Find potential hallucinations in transcriptions
- Find repeated phrase patterns
- Count BLANKAUDIO tokens
- Find low-confidence + short transcriptions (likely noise)
- Tuning Scenarios
- Aggressive Hallucination Filtering
View current phrases
grep -A20 "silenceHallucinationPhrases:" /Users/thw/Projects/WaveCap/state/config.yaml
grep -E "(hallucination|repetition|blank)" /Users/thw/Projects/WaveCap/state/config.yaml
curl -s http://localhost:8000/api/transcriptions/export | \
jq '[.[] | select(
jq '[.[] | select(.text | test("(\\b\\w{4,}\\b).*\\1.*\\1.*\\1"))] |
jq '[.[] | select(.text == "[BLANK_AUDIO]")] | length'
jq '[.[] | select(.confidence < 0.6 and (.text | length) < 20)] |
launchctl stop com.wavecap.server && sleep 2 && launchctl start com.wavecap.server
cd /Users/thw/Projects/WaveCap/backend && source .venv/bin/activate && python3 << 'EOF'What does the wavecap-hallucination skill do?
Configure WaveCap hallucination detection and prevention. Use when Whisper outputs gibberish, repeated phrases, or phantom text on silent audio.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill wavecap-hallucination-tobiaswooldridge-wavecap --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.