v0.4.1 - Add Icon

Add Icon

v0.4.0 - Remove system tray

Changes

• Remove system tray: The minimize-to-tray feature was non-functional due to an eframe/winit limitation. The app now stays in the taskbar like a normal window.

Download

Download whisper-native.exe and run it directly. Models are downloaded automatically on first launch.

v0.3.0 - Fix Large V3 transcription

Changes

• Fix Large V3 transcription: Large V3 was always outputting English with a leading comma regardless of language selection. Root cause: Large V3 has 100 language tokens (adds Cantonese) which shifts special token IDs by +1. The model was receiving <|translate|> instead of <|transcribe|> and <|nospeech|> instead of <|notimestamps|>. Token IDs are now computed dynamically per model variant.
• Remove Turbo variant: Whisper Large V3 Turbo has been removed from the application. Only Medium and Large V3 are available.

Download

Download whisper-native.exe and run it directly. Models are downloaded automatically on first launch.

v0.2.0 - Inline Settings & Hotkey Overhaul

Changes

• Inline settings: Settings moved from overlay to main screen (no more popup)
• Hotkey system rewrite: Replaced global_hotkey crate with GetAsyncKeyState polling
  • Supports modifier-only combos (e.g. Ctrl+Win)
  • Interactive "Change" button to capture any key combination
  • More reliable press/release detection
• System tray: Close button hides to tray instead of quitting
• No console window: Hidden via #![windows_subsystem = "windows"]
• UI cleanup: Title + status on same line, larger hotkey instruction text

Download

Download whisper-native.exe and run it directly. Models are downloaded on first launch.

v0.1.0 - Whisper Burn Initial Release

Whisper Burn v0.1.0

Native Rust implementation of OpenAI's Whisper speech recognition with GPU acceleration.

Features

• Whisper Large V3 & V3 Turbo support with Q4_0 quantized GGUF models
• GPU-accelerated inference via wgpu (Vulkan/Metal/DirectX)
• Custom WGSL compute shaders for fused Q4 dequantization + matrix multiplication
• Push-to-Talk with global hotkey and system tray integration
• 99 languages supported for transcription
• Native desktop app built with eframe/egui

Downloads

• whisper-native.exe — Windows desktop application

Usage

1. Download whisper-native.exe
2. Run the application — it will automatically download the required model on first launch
3. Use Push-to-Talk or record audio to transcribe speech to text