v2.0.0 — Open Source
  ███╗   ███╗██╗  ██╗███████╗    ██████╗
  ████╗ ████║██║  ██║██╔════╝    ██╔══██╗
  ██╔████╔██║███████║███████╗    ██║  ██║
  ██║╚██╔╝██║╚════██║╚════██║    ██║  ██║
  ██║ ╚═╝ ██║     ██║███████║    ██████╔╝
  ╚═╝     ╚═╝     ╚═╝╚══════╝    ╚═════╝

Download anything.
Extract every voice.

A Qt6 desktop app that downloads from 1,000+ sites and isolates vocals using Meta's Demucs AI — all offline-first, bilingual, dark/light themed.

1000+
Supported Sites
4
Audio Formats
AI
Demucs htdemucs_ft
2
Languages (AR/EN)

Everything in one
offline-first app

No cloud accounts, no subscriptions. Tools are cached locally so you're never re-downloading.

🌐
1,000+ Sites
Powered by yt-dlp. YouTube, SoundCloud, Twitter/X, TikTok, Instagram, Vimeo, Dailymotion and hundreds more.
yt-dlp
🎤
AI Vocal Extraction
Meta's htdemucs_ft model separates vocals from instruments. Two isolation levels: Normal (shifts=2) and Extreme (shifts=4).
Demucs AI
📁
Local File Support
Drag & drop or browse any video or audio file. ffmpeg auto-extracts audio from videos before AI processing.
ffmpeg
🧹
Auto Cleanup
After processing, the app automatically deletes the original download, extracted WAV, and demucs temp folder to save disk space.
smart pipeline
🌙
Dark / Light Themes
Toggle between a deep dark code-editor theme and a clean light mode. Preference is saved between sessions via QSettings.
Qt Fusion
🌐
Bilingual UI
Full Arabic (RTL) and English (LTR) toggle. The entire layout direction flips with one click — no restart needed.
AR / EN
Offline-First Tools
yt-dlp and ffmpeg are downloaded once to your local app data folder. Subsequent runs use cached binaries — no internet needed for processing.
local cache
🖥
GPU Acceleration
CUDA mode accelerates Demucs on NVIDIA GPUs. Automatic CPU fallback if CUDA is unavailable or fails.
CUDA / CPU

Every format you need

Convert audio and download video in the quality you choose.

🎵
Audio Output
MP3 (VBR q2) M4A (AAC 256k) WAV (PCM) FLAC (lossless)
🎬
Video Download
Best Quality 4K (2160p) 1080p 720p 480p
📥
Local Input Files
MP3 M4A WAV FLAC OGG OPUS MP4 MKV WEBM AVI MOV
🤖
AI Model
htdemucs_ft shifts=2 (normal) shifts=4 (extreme) --two-stems=vocals CUDA / CPU

How it works

From URL to clean vocal track in one click.

INPUT ┌──────────────────┐ ┌──────────────────────────────────────────┐ URL (yt-dlp) │────▶│ Download audio as WAV Local File │────▶│ Extract audio from video (ffmpeg -vn) └──────────────────┘ └────────────────────┬─────────────────────┘ ┌──────────────────────────────────────────────▼─────────────────────┐ "No Music" checked? YES ──▶ demucs -n htdemucs_ft --two-stems=vocals (shifts=2 normal / shifts=4 + overlap=0.25 extreme) ──▶ vocals.wav NO ──▶ Skip demucs └──────────────────────────────────────────────┬─────────────────────┘ ┌──────────────────────────────────────────────▼─────────────────────┐ ffmpeg convert ──▶ MP3 / M4A / WAV / FLAC Output: "Title (no music).ext" or "Title.ext" └──────────────────────────────────────────────┬─────────────────────┘ ┌──────────────────────────────────────────────▼─────────────────────┐ 🧹 AUTO CLEANUP Delete: original download · temp WAV · demucs output folder └─────────────────────────────────────────────────────────────────────┘

Get up and running

One script handles everything — from dependencies to binary installation.

1
Clone the repository
git clone https://github.com/mahmoudelsheikh7/M4S_D cd M4S_D
2
Run the installer
Auto-detects pacman / apt / dnf / zypper. Installs Qt6, ffmpeg, yt-dlp, demucs.
chmod +x install_linux.sh ./install_linux.sh # do NOT run as root
3
Launch
m4s_d # or open "m4s d" from your app menu
💡 Supported: Arch/Manjaro · Ubuntu/Debian/Mint · Fedora/RHEL/Rocky · openSUSE
Note: Demucs downloads PyTorch (~1–3 GB), allow 5–20 min.
1
Allow script execution (once)
Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned
2
Run the Windows installer
Downloads yt-dlp.exe, ffmpeg.exe and sets up demucs in a Python venv.
.\install_windows.ps1
3
Build the GUI
Open Qt Creator → File → Open → select CMakeLists.txt → Build → Run.
Or with MSYS2:
mkdir build && cd build cmake .. -G "MinGW Makefiles" -DCMAKE_BUILD_TYPE=Release mingw32-make -j4
Build manually on any platform with Qt6 and CMake installed.
# Clone git clone https://github.com/mahmoudelsheikh7/M4S_D && cd M4S_D # Configure mkdir build && cd build cmake .. -DCMAKE_BUILD_TYPE=Release # Build cmake --build . --parallel # Install (Linux, optional) sudo cmake --install .
Requirements: CMake ≥ 3.16 · Qt6 (Core, Gui, Widgets, Network) · C++17 compiler
💛
Support m4s d

m4s d is free and open source. If it saves you time or you find it useful, a crypto donation keeps the project alive and motivates future features.

USDT · TRC20 · TRON Network
Wallet Address
TLTRj89qEp1oPcEQjAsqiDWutd1ezd4W4y
Only send USDT on TRC20 (TRON) network. Always verify the address.