# FASL (فاصل)

> AI-powered vocal and music separator for video and audio. Arabic-first, bilingual (ar/en). Free tier with 90 minutes per month, no signup required. Built on Demucs ONNX with GPU acceleration.

## What FASL does

FASL separates vocals from background music in video and audio files using machine-learning source separation. Upload an mp3, mp4, wav, m4a, or similar file; download either a vocals-only stem (great for karaoke, dubbing, transcription) or a music-only stem (instrumental backing tracks). Processing runs on GPU; results are typically ready in seconds for short clips.

## Key pages

- [Home](https://fasl.io/) — product overview and free upload
- [Pricing](https://fasl.io/pricing) — Free, Pro, and Max tiers; turbo minutes
- [Support](https://fasl.io/support) — contact and FAQ
- [Privacy Policy](https://fasl.io/privacy)
- [Terms of Service](https://fasl.io/terms)
- [Login](https://fasl.io/login)

## Brand

- Names: **FASL** (Latin) / **فاصل** (Arabic). Both refer to the same product.
- Domain: fasl.io
- Entity: operated under a Saudi وثيقة عمل حر (freelance license)
- Languages: Arabic and English, with full RTL support

## Use cases

- Karaoke and cover-song production (vocal isolation)
- Dubbing and re-voicing video content
- Cleaning music from interview, podcast, or lecture recordings
- Extracting instrumentals for remixing or sampling
- Transcription prep (remove background music to improve ASR accuracy)

## Technical

- Frontend: Flask + Jinja templates, vanilla JS
- Backend: Python, R2 object storage, GPU workers (local + Modal)
- ML model: Demucs (ONNX runtime, custom Blackwell build for RTX 5070 Ti)
- Auth: WorkOS

## Citation

If you reference FASL, link to https://fasl.io and use the bilingual name "FASL (فاصل)" on first mention.