Audio Cleaner

A cross‑platform desktop application that removes background noise from any video’s audio track using the [DeepFilterNet 3] neural noise‑suppression model and an intuitive PyQt 6 interface.
It extracts the soundtrack, denoises it with DeepFilterNet, then remuxes the cleaned audio back into the original video with FFmpeg – all in one click.

---

## Key Features

• Point‑and‑click GUI built with PyQt 6 – no command line needed.
• Preview player – play the source or cleaned video inside the app.
• Adjustable attenuation limit slider (1‑60 dB) or a safe recommended default.
• Progress bar & status messages courtesy of a background QThread worker.
• Automatic remux step strips problematic metadata (e.g. Sony rtmd) before processing.
• Works with MP4, MOV, AVI, MKV and any format FFmpeg can decode.
• Cross‑platform: Windows, macOS, Linux (tested on Python 3.9+).
• Clean shutdown & temp‑file cleanup even on errors.

---

## Quick Start

# 1 – Clone this repository
git clone https://github.com/your‑username/audio-cleaner.git
cd audio-cleaner

# 2 – Create & activate a virtual environment (optional but recommended)
python -m venv .venv
source .venv/bin/activate   # On Windows: .venv\Scripts\activate

# 3 – Install dependencies
pip install -r requirements.txt

# 4 – Download the DeepFilterNet 3 model (≈ 30 MB)
python -m df --download-models

# 5 – Run the app
python deep.py

The first launch will prompt you to select an input video and a save location.
Click “Denoise Video” and let the progress bar reach 100 %. Voilà – background noise gone!

---

## Requirements

Package	Tested Version	Notes
Python	3.9 – 3.12
[deepfilternet]	≥ 0.4
PyQt6	≥ 6.5	GUI, multimedia widgets
moviepy	≥ 1.0	Video I/O
soundfile	≥ 0.12	WAV read/write
imageio‑ffmpeg	≥ 0.4	FFmpeg binary download helper
Torch
TorchAudio

A pre‑filled requirements.txt is provided for your convenience.

---

## How It Works

           ┌────────────┐ 1. Extract audio with MoviePy
Input MP4 ─► Remux step ├──────────┐
           └────────────┘          │
                                   ▼
                            2. Denoise WAV
                               with DF‑Net
                                   ▼
           ┌────────────┐ 3. Combine audio & video with FFmpeg
Clean MP4 ◄─┤  FFmpeg   │
           └────────────┘

The heavy lifting is done in DenoiseWorker (see deep.py) running in its own QThread so the UI stays responsive.

---

## Packaging a Standalone Executable (optional)

Install PyInstaller, then:

pyinstaller --noconfirm --windowed --onefile \
  --add-data "models/DeepFilterNet3:models/DeepFilterNet3" \
  --add-binary "$(python - <<'PY'
import imageio_ffmpeg, sys, os; print(imageio_ffmpeg.get_ffmpeg_exe()+os.pathsep+'ffmpeg')
PY)":. \
  deep.py

The resulting dist/deep (or deep.exe on Windows) can be shipped without requiring Python on the end‑user’s machine.

---

## Troubleshooting

Symptom	Possible Cause / Fix
“Core dependencies missing” in status bar	pip install -r requirements.txt
No audio in output	Set console=true in VideoDenoiser.spec when creating an executable
FFmpeg executable not found	pip install imageio‑ffmpeg or add FFmpeg to PATH.
Crash when loading video	Some raw camera files have invalid edit lists – the built‑in remux step usually fixes this.

---

## License

Distributed under the MIT License – see LICENSE for details.

---

## Acknowledgements

• DeepFilterNet by A. Rosenkranz et al.
• MoviePy – video editing in Python
• FFmpeg – the Swiss‑Army knife of video processing