Repository navigation
Murmur v0.1.1
What's New
- NVIDIA Parakeet TDT v3 — Replaced faster-whisper with NeMo Parakeet TDT v3 for higher accuracy multilingual transcription with built-in timestamps
- Two model sizes: Parakeet TDT 0.6B (~700 MB) and Parakeet TDT 1.1B (~1.3 GB)
- VRAM-aware model recommendations — auto-selects the best model for your GPU
- Drag & drop audio files for transcription
- SRT & JSON export alongside Markdown
- Auto-update notifications via electron-updater
- Live transcription improvements (3s interval, deduplication)
- Reliability: per-request timeouts, auto-restart on backend crash, streaming checksums for >2GB downloads
- Deferred summary model loading — saves ~2-3 GB VRAM on startup
Install
Download and run Murmur-Setup-0.1.1.exe. The ML backend (~5 GB) downloads automatically on first launch.