v0.55.1 v0.2.0
imbhargav5/open-recorderOct 2, 2026 → Oct 2, 20260 days between releases
What's New
Introduces Nemotron 3 diarization, MagpieTTS v2607, and upgraded llama.cpp with performance optimizations
Breaking Changes
- Model index updated to point to MagpieTTS v2607, potentially breaking existing configurations
- llama.cpp submodule upgraded, which may require code adjustments
New Features
- Support for Nemotron 3 high-resolution diarization
- MagpieTTS v2607 with Arabic, Korean, and Portuguese support
- Nemotron Labs VoiceChat for real-time speech-to-speech interaction
- Live diarized transcription with speaker turns and endpointing
- Text normalization (ITN and TN) support in archives
Improvements
- Faster MagpieTTS performance on CUDA and CPU via persistent decoder kernels
- Streaming ASR improvements with context retention across endpoints
- Upgraded llama.cpp to a single pinned submodule
- Nightly builds available via --channel nightly flag
Upgrade Notes
The model index now points to MagpieTTS v2607. The llama.cpp submodule has been upgraded to a single pinned version. Note that Nemotron Labs VoiceChat is CUDA-only and must be built from source.
The Two Releases
v0.55.1
Oct 2, 2026· 20 days ago
This release optimizes the `pipecat-bot-py` example by switching the default Text-to-Speech provider to Deepgram Aura for faster response times. It also resolves a pause-mode barge-in issue where bot audio was incorrectly dropped after a resolution.
Full release notesv0.2.0
Oct 2, 2026· 20 days ago
This release introduces support for high-resolution diarization models like Nemotron 3 and MagpieTTS v2607, alongside major performance optimizations for CUDA and CPU backends. It also adds the Nemotron Labs VoiceChat feature for real-time speech-to-speech interaction and upgrades the underlying llama.cpp library.
Full release notes