Self-hosted media downloader for YouTube, TikTok, Instagram & Facebook.
Paste a link, preview it, pick a format — with a live job dashboard, waveform audio trimming, and Lalal.ai stem separation.

Dashboard — paste a link, preview the media, pick a format, and track every job live.

Authenticated login with invisible, server-verified anti-bot protection.

Why fetchly

• A web UI, not the CLI — no yt-dlp flags to remember; format and quality are picked from menus.
• It stays running — a persistent dashboard for queued, active, done, and failed jobs, not a one-shot command.
• More than downloading — trim audio on a waveform and split vocals from instrumentals without leaving the app.
• Self-hosted — a single container, SQLite on a mounted volume, authenticated access, and a hardened default setup.

Features

Supported platforms

Video and audio downloads from YouTube, TikTok, Instagram, and Facebook.

Note: Only publicly accessible URLs work — private videos will not download.

Quickstart

Docker is the recommended way to run fetchly. The image (Docker Hub) bundles ffmpeg, yt-dlp, yt-dlp-ejs, and deno — nothing else to install.

Docker Compose

# compose.yaml
services:
  fetchly:
    image: giiibates/fetchly:latest
    container_name: fetchly
    restart: unless-stopped
    ports:
      - "8000:8000"
    environment:
      FETCHLY_SECRET_KEY: ${FETCHLY_SECRET_KEY:?generate with: openssl rand -base64 32}
      FETCHLY_ADMIN_PASSWORD: ${FETCHLY_ADMIN_PASSWORD:?the admin login password}
    volumes:
      - ./data:/app/data

export FETCHLY_SECRET_KEY="$(openssl rand -base64 32)"
export FETCHLY_ADMIN_PASSWORD="change-me"
docker compose up -d

A fuller compose file with logging, timezone, and reverse-proxy notes lives in docker/docker-compose.yml.

docker run

docker run --rm \
  -p 8000:8000 \
  -e FETCHLY_SECRET_KEY="$(openssl rand -base64 32)" \
  -e FETCHLY_ADMIN_PASSWORD="change-me" \
  -v "$PWD/data:/app/data" \
  giiibates/fetchly:latest

Then open http://127.0.0.1:8000 and sign in as admin.

To build the image yourself: docker build -f docker/Dockerfile -t fetchly .

Standalone

Requires Linux, Python 3.13, and ffmpeg, yt-dlp, deno on PATH.

pip install -r requirements.txt && pip install yt-dlp yt-dlp-ejs
export FETCHLY_SECRET_KEY="$(openssl rand -base64 32)"
export FETCHLY_ADMIN_PASSWORD="change-me"
python run.py

Configuration

Everything is set through environment variables (with Compose, put them in an .env file next to the compose file).

Lalal.ai stem separation is enabled by entering an API key under Settings in the app — no environment variable required.

Reverse proxy

fetchly speaks plain HTTP inside the container. Terminate TLS at a proxy (Caddy, nginx, Traefik), forward to port 8000, and set FETCHLY_BEHIND_HTTPS=1. An example Caddyfile is included.

Built with

• yt-dlp + yt-dlp-ejs and deno — media extraction and YouTube JS-challenge solving
• ffmpeg — transcoding, audio trimming, and waveform generation
• wavesurfer.js — browser-based waveform display and trim UI
• Gunicorn + Uvicorn — ASGI server used in Docker
• Lalal.ai — optional vocals/instrumental stem separation

The image ships upstream static ffmpeg/ffprobe builds and stable yt-dlp PyPI releases; Debian's media packages are pinned out so they cannot shadow them.