Release v1.9.1
Release v1.9.1
feat: VOICE_RECOGNITION STT audio source, ambient [SOUND] tag filtering,
HTML sanitization, and concurrent credits scroll
- Switch Live Mode AudioRecord source to VOICE_RECOGNITION in AudioCaptureService.kt for high quality uncompressed 16kHz speech.
- Filter out [SOUND], [Music], [NOISE] ambient tags and silence hallucinations in SherpaSTTEngine.kt.
- Increase VAD silence duration to 950ms in LiveVoiceOverlay.kt to prevent mid-sentence word duplication.
- Sanitize
and
HTML tags into newlines in MarkdownText.kt for clean chat bubble formatting. - Run EndCredits volume fade concurrently so scroll begins at t=0 immediately without initial delay, with 28f crawl speed.
- Bump version to 1.9.1 (versionCode 60).