Voice Prompt Cleanup

A GUI application for preprocessing audio files for speech-to-text (STT) workflows. Optimizes audio before sending to multimodal APIs like Google Gemini or ASR models.

Features

• PyQt6 GUI with drag-and-drop support
• Batch processing - process single files, multiple files, or entire folders
• Persistent settings - remembers your output folder preference
• Safe processing - outputs as <filename>_processed.mp3 to never overwrite originals

Audio Processing Pipeline

1. Convert to Mono - STT models don't need stereo
2. Downsample to 16kHz - matches what most APIs use internally
3. Speech EQ - 80Hz-8kHz bandpass filter for voice clarity
4. Gentle Compression - evens out speech dynamics
5. Truncate Silences - removes long pauses
6. Normalize Audio - consistent levels without clipping
7. Export as MP3 - compressed format suitable for API upload

Installation

From Debian Package (Recommended)

# Clone the repository
git clone https://github.com/danielrosehill/Voice-Prompt-Cleanup-Script.git
cd Voice-Prompt-Cleanup-Script

# Build the package
./build-deb.sh

# Install
sudo apt install ./build/voice-prompt-cleanup_1.0.0-1.deb

Dependencies

If installing manually, you need:

• Python 3
• PyQt6 (pip install PyQt6 or sudo apt install python3-pyqt6)
• ffmpeg (sudo apt install ffmpeg)

Running Without Installation

# Install dependencies
sudo apt install python3-pyqt6 ffmpeg

# Run directly
./voice_prompt_cleanup_gui.py

Usage

GUI Application

Launch from your application menu as "Voice Prompt Cleanup" or run:

voice-prompt-cleanup

1. Add files by dragging them onto the window, or use "Add Files..." / "Add Folder..."
2. Set output folder (optional) - enable custom output folder to save all processed files to one location
3. Click Process Files

Command Line (Script Only)

./process_audio.sh input.mp3 [output.mp3]

Updating

To update to the latest version:

cd Voice-Prompt-Cleanup-Script
./update-package.sh

This will pull the latest changes, rebuild, and reinstall the package.

Supported Formats

Input: MP3, WAV, FLAC, OGG, M4A, AAC, WMA, OPUS, WEBM, MP4, MKV, AVI, MOV

Output: MP3 (64kbps, 16kHz mono)

Configuration

Settings are stored in ~/.config/voice-prompt-cleanup/settings.json:

• Output folder path
• Whether to use custom output folder
• Last used input folder

Target Use Case

Primary target: Google Gemini Audio Understanding and similar multimodal APIs

• Accepts: MP3, WAV, FLAC, OGG, etc.
• Typically downsamples to 16kHz internally
• Often has file size limits (e.g., 20MB for Gemini)

The preprocessing optimizes for these constraints while maintaining speech quality.

License

MIT License - see LICENSE for details.