Not Whisper Flow

100% local, free, open-source voice-to-prompt tool. Speak your coding problem, get a polished prompt for AI assistants. Or use it as a voice notes dictionary.

What It Does

Two modes:

1. Code Prompt Mode - Speak a rough description of what you want to build/fix/debug. The app transcribes it with Whisper and uses a local SLM to turn it into a clean, structured prompt you can paste into Claude, ChatGPT, Cursor, etc.

2. Voice Notes Mode - Speak to save notes, docs, and knowledge snippets. Builds a searchable local knowledge base from your voice.

Everything runs locally. No API keys. No cloud. No cost.

Requirements

• Python 3.8+
• FFmpeg (for Whisper)
• ~6 GB disk space (for models)

Quick Start

# Install dependencies
pip install -r requirements.txt

# First-time setup (downloads models)
python -m utils.installer

# Run
python main.py

Or on Windows, just double-click start.bat.

Usage

1. Press Ctrl+Shift+Space to start recording
2. Speak your coding problem or note
3. Press Ctrl+Shift+Space again to stop
4. Preview the enhanced prompt in the overlay
5. Press Enter to accept and auto-type, or Esc to cancel

Switch between Code Prompt and Voice Notes mode via the system tray menu.

Models

Model	Size	Speed	Use Case
Whisper base	~140 MB	Fast	Speech-to-text (default)
Whisper tiny	~75 MB	Fastest	Lower accuracy, faster
Qwen2.5-0.5B	~1 GB	Good on CPU	Prompt enhancement (default)
Qwen2.5-1.5B	~3 GB	Slower on CPU	Better quality prompts
SmolLM2-360M	~720 MB	Fastest on CPU	Lightest option

Configuration

Settings are stored in ~/.whisper_flow/config.json. Key options:

Setting	Default	Options
app_mode	code_prompt	code_prompt, voice_notes
whisper_model	base	tiny, base, small, medium, large
slm_model	qwen2.5-0.5b	smollm2-360m, qwen2.5-0.5b, qwen2.5-1.5b
enhancement_mode	preview	auto, preview, off
hotkey_toggle_recording	ctrl+shift+space	Any key combo

Project Structure

not_whisper_flow/
├── main.py              # App orchestrator
├── config.py            # JSON-persisted settings
├── audio/               # Mic capture, preprocessing, VAD
├── transcription/       # Whisper STT engine
├── slm/                 # Local SLM prompt enhancement
├── notes/               # Voice notes storage
├── automation/          # Hotkeys + auto-typing
├── ui/                  # System tray + overlay window
└── utils/               # Logger + installer

License

Open source. Free to use.