termux-stt

Version Archive

Release history and upgrade guides for termux-stt.

v1.0.0 (Initial Public Release) - 2026-08-20

• Unified Multi-Engine Architecture: Single create_engine() API for whisper.cpp, vosk, and sherpa-onnx.
• Pure-Python Speaker Diarization: Custom K-Means and Cosine Distance matrix without numpy/sklearn dependencies.
• Hybrid Diarization Pipeline: Vosk X-Vector + Whisper STT under 1.5 GB RAM.
• Audio Pipeline: ffmpeg 16kHz mono converter, Silero-VAD integration, and Termux microphone recorder.
• Model Hub: Automatic model downloader with SHA-256 verification and cache management.
• Export Standards: SRT, WebVTT, NIST RTTM, and structured JSON output.
• CLI Suite: transcribe, listen, diarize, models, doctor, and benchmark.
• Dual-Engine Support: First-class Node.js npm package with identical capabilities.
• Live Showcase Page: Interactive web audio playback and synchronized SRT subtitle timeline.