@tychenjiajun/pi-voice 0.6.0 → 0.8.0

package/README.md

@@ -64,7 +64,7 @@ After ASR produces a transcript, Pi Voice can refine it with the same model (and
 Set it up in `/voice-settings` → **Post-processing** → **On**. You then choose the model:
 
 - **Pi's current model** (default) — whichever model the agent is using, exactly like compaction;
-- **a different model** — any model from Pi's registry, using that provider's configured authentication.
+- **a different model** — any model from Pi's registry, using that provider's configured authentication. The picker is searchable: type a keyword (a model name or part of `provider/id`) to filter the list, `Enter` selects the highlighted match, and `Esc` clears the search.
 
 Pi Voice makes a small nested request through Pi's model registry, so it automatically reuses Pi's model definitions and API keys (or OAuth). No base URL, key, or timeout is stored in Pi Voice settings.
 
@@ -91,7 +91,7 @@ pi install npm:@tychenjiajun/pi-voice
 
 ## File transcription and FFmpeg
 
-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
+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
 
 ```bash
 # macOS with Homebrew
@@ -116,6 +116,36 @@ When FFmpeg is unavailable, `transcribe_file` reports platform-specific guidance
 
 ## Changelog
 
+### 0.8.0
+
+**Added — Upstream sync: voice activity detection, long-audio chunking, and language detection**
+
+- Synced with upstream `@earendil-works/pi-voice` 0.2.0. The remote ASR backend and compaction-style post-processing from this fork are unchanged.
+- **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`).
+- **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.
+- **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.
+- **Chinese output** now keys off the transcript text (Han characters, no kana) instead of language detection, so remote and short-clip transcripts convert reliably.
+- **New catalog models**: Parakeet Ultra and Granite Speech 5.0 470M TurboCTC.
+- Decoded audio for `transcribe_file` is raised to 512 MiB (about 2.3 hours).
+
+**Fixed**
+
+- Whisper-family out-of-memory failures during transcription now retry on the CPU device instead of failing the take.
+
+**Changed**
+
+- Dependencies: `transcribe-cpp` 0.4.0 and `@handy-computer/recorder` 0.2.1.
+- Catalog metadata is imported from `catalog/catalog.json` directly; the generated `catalog.generated.ts` copy is gone.
+- Built against `@earendil-works/pi-coding-agent` 1.3.0.
+
+### 0.7.0
+
+**Added — Searchable post-processing model picker**
+
+- `/voice-settings` → **Post-processing** → model now has a search box, so a registry with a thousand-plus models can be narrowed by typing instead of arrowing through the whole list.
+- The search uses the same fuzzy matcher as the transcription model picker (`fuzzyFilter` from `@earendil-works/pi-tui`) and matches the model name as well as `provider/id`, so `wspr` finds `openai/whisper-1` and `claude` finds `anthropic/claude-opus-4-5`.
+- Results are sorted by match quality with the cursor on the best match (Enter picks it), `Esc` clears the search and restores the full list, and the subtitle reminds you that typing filters the models.
+
 ### 0.6.0
 
 **Changed — Fuzzy model search**