@tychenjiajun/pi-voice 0.7.0 → 0.8.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (33) hide show
  1. package/README.md +23 -1
  2. package/catalog/catalog.json +1501 -0
  3. package/catalog/recommendations.json +67 -27
  4. package/models/README.md +25 -0
  5. package/models/silero-vad-v6.2-F32.gguf +0 -0
  6. package/package.json +8 -6
  7. package/src/audio-constants.ts +2 -0
  8. package/src/audio.ts +2 -5
  9. package/src/catalog.ts +6 -2
  10. package/src/chinese.ts +17 -6
  11. package/src/chunking.ts +319 -0
  12. package/src/dictation-controller.ts +61 -18
  13. package/src/file-audio.ts +2 -2
  14. package/src/file-transcription.ts +6 -3
  15. package/src/language-detection-policy.ts +127 -0
  16. package/src/languages.ts +4 -3
  17. package/src/model-activation.ts +50 -4
  18. package/src/model-cells.ts +28 -6
  19. package/src/model-picker.ts +29 -12
  20. package/src/model-selection-controller.ts +14 -9
  21. package/src/models.ts +15 -1
  22. package/src/onboarding.ts +13 -1
  23. package/src/pcm-chunker.ts +14 -4
  24. package/src/recommendation-picker.ts +9 -7
  25. package/src/recommendations.ts +4 -2
  26. package/src/runtime.ts +9 -1
  27. package/src/settings-menu.ts +40 -9
  28. package/src/settings.ts +35 -3
  29. package/src/transcription-service.ts +205 -37
  30. package/src/transcription.ts +72 -24
  31. package/src/try-it.ts +11 -2
  32. package/src/vad.ts +301 -0
  33. package/src/catalog.generated.ts +0 -1305
package/README.md CHANGED
@@ -91,7 +91,7 @@ pi install npm:@tychenjiajun/pi-voice
91
91
 
92
92
  ## File transcription and FFmpeg
93
93
 
94
- The agent can call `transcribe_file` for local audio or video files. Decoded audio is limited to 128 MiB (about 35 minutes). File decoding requires the `ffmpeg` executable. Install FFmpeg with your system package manager if you don't already have it installed
94
+ The agent can call `transcribe_file` for local audio or video files. Decoded audio is limited to 512 MiB (about 2.3 hours). File decoding requires the `ffmpeg` executable. Install FFmpeg with your system package manager if you don't already have it installed
95
95
 
96
96
  ```bash
97
97
  # macOS with Homebrew
@@ -116,6 +116,28 @@ When FFmpeg is unavailable, `transcribe_file` reports platform-specific guidance
116
116
 
117
117
  ## Changelog
118
118
 
119
+ ### 0.8.0
120
+
121
+ **Added — Upstream sync: voice activity detection, long-audio chunking, and language detection**
122
+
123
+ - Synced with upstream `@earendil-works/pi-voice` 0.2.0. The remote ASR backend and compaction-style post-processing from this fork are unchanged.
124
+ - **Voice activity detection** using a bundled Silero model (`models/silero-vad-v6.2-F32.gguf`): silence and noise are filtered from dictation so only speech is transcribed. The hidden `voiceActivityDetection` setting in `settings.json` turns it off (default `true`).
125
+ - **Long-audio chunking**: clips that do not fit one model window are cut at pauses before batch transcription, so non-Whisper models no longer truncate long recordings. The hidden `longformChunking` setting turns it off (default `true`); models that handle long audio natively (Whisper, Parakeet Ultra) and streaming models skip it.
126
+ - **Automatic language detection**: when a multilingual model has no native detection and Auto is selected, the shared **VoxLingua107** detector picks the language. Download it from the language picker; Auto is only offered when the model supports at least two of your preferred languages and the detector can identify them.
127
+ - **Chinese output** now keys off the transcript text (Han characters, no kana) instead of language detection, so remote and short-clip transcripts convert reliably.
128
+ - **New catalog models**: Parakeet Ultra and Granite Speech 5.0 470M TurboCTC.
129
+ - Decoded audio for `transcribe_file` is raised to 512 MiB (about 2.3 hours).
130
+
131
+ **Fixed**
132
+
133
+ - Whisper-family out-of-memory failures during transcription now retry on the CPU device instead of failing the take.
134
+
135
+ **Changed**
136
+
137
+ - Dependencies: `transcribe-cpp` 0.4.0 and `@handy-computer/recorder` 0.2.1.
138
+ - Catalog metadata is imported from `catalog/catalog.json` directly; the generated `catalog.generated.ts` copy is gone.
139
+ - Built against `@earendil-works/pi-coding-agent` 1.3.0.
140
+
119
141
  ### 0.7.0
120
142
 
121
143
  **Added — Searchable post-processing model picker**