শব্দ-হটাও · ShobdoHotao

আপনার রেকর্ডিং থেকে আওয়াজ দূর করুন — সম্পূর্ণ অফলাইনে। Remove background noise from your recordings — fully offline.

Select a noisy recording, choose a strength, click Clean Audio / আওয়াজ দূর করুন, compare, and export. No account. No upload. No limit.

Works on audio files (MP3/WAV/M4A/FLAC/…) and on the audio track of videos (MP4/MOV/MKV/AVI/WebM): the video is copied losslessly and only its noisy audio is cleaned and remuxed back. See docs/PRODUCT_SPEC.md.

---

⬇️ Download & install (Windows)

No Python, account, or setup needed — just download and run:

1. Open the Releases page and download the latest ShobdoHotao-Setup.exe.
2. Run it. Windows may warn "Windows protected your PC" (the app isn't code-signed) — click More info → Run anyway. This is normal and safe.
3. Click through the installer; it adds a desktop shortcut.
4. Launch ShobdoHotao. On the first run only, it downloads the AI model (~80 MB) once — after that it works fully offline. Your audio never leaves your computer.

---

Privacy first

• Audio never leaves your computer. No login, API key, cloud, telemetry, ads, or quota.
• Core processing works without internet after installation.
• App source is MIT; third-party licenses are preserved in THIRD_PARTY_NOTICES.md.

Status

Working desktop app (audio + video) with a managed local library. Cleaned files are saved under your Documents folder:

Documents/ShobdoHotao/
├── Cleaned Files/{Audio,Video}/   # your results
├── Database/  Logs/  Temp/

During processing a dedicated Processing view shows a live pipeline stepper, a reactive signal visualizer, and a real engine log (genuine FFmpeg progress for video; honest elapsed/stage for AI denoising — no fake percentages). After processing you name the file (live Windows-safe validation), then a completion screen offers Play/Open, Show in Folder, and Clean Another. A Cleaned Files button opens your library. Everything stays offline.

See docs/ARCHITECTURE.md for layer boundaries, docs/PRODUCT_SPEC.md for behaviour, and docs/ROADMAP.md for what's next (in-app library screen, storage settings, installer).

Requirements

• Windows 10/11 x64 (default target), CPU-only.
• Python 3.11 (64-bit) for development.

PyTorch pin: DeepFilterNet 0.5.6 requires torch==2.0.1 + torchaudio==2.0.2 (newer torchaudio removed torchaudio.backend.common, which DeepFilterNet imports). setup_windows.ps1 installs these CPU builds for you.

Setup (development)

# from the repo root
./scripts/setup_windows.ps1     # creates .venv, installs CPU PyTorch + deps
./scripts/run_windows.ps1       # launches the app

Run tests

The pure-Python core has no heavy dependencies and tests run anywhere:

python -m pytest          # or: PYTHONPATH=src python -m pytest
ruff check .

Build (Windows release ZIP)

./scripts/build_windows.ps1     # folder-based PyInstaller build

License

MIT — see LICENSE.