Qevoxyl Audio Transcription System — ASR / STT
SOFTWARE |
333
Voice in. Clean text out.
A privacy-first Windows desktop app for automatic speech recognition (ASR) / speech-to-text (STT). Qevoxyl runs Whisper offline, ships with an FFmpeg shim, and offers a clean CustomTkinter UI with optional Online TTS playback. Built for per-user installs (no UAC), with right-click Explorer integration and an AV-friendly one-folder packaging approach.
- Offline Whisper (PyTorch) with local model caching and GPU fp16 when available.
- FFmpeg hardening: resolves via imageio-ffmpeg; local fallback copy to
models/ffmpeg-bin. - Explorer context menu: “Transcribe with Qevoxyl” for
.wav/.mp3. - Per-user install: Inno Setup to
%LOCALAPPDATA%\Qevoxyl(no admin/UAC). - One-folder PyInstaller build with version info + manifest; UPX disabled for stability.
- Session guard & reset: cancel in-flight work, start new session, and keep the UI responsive.
- Persistent settings: appearance, model selection, and Online TTS toggle in
settings.json.
