8mb.local – Video Compressor

Caution

IMPORTANT DISCLAIMER: This software is provided "as is" without warranty of any kind.

This is a heavily modified fork of https://github.com/JMS1717/8mb.local

Whilst hardware support has been kept the same, with native ARM64 and MacOS support built up, outside of this scope, it has not been tested directly with Nvidia/AMD/Intel

Support will not be provided, as it has been purpose built to run in a Container with a native Dameon in MacOS for Hardware Accel.

8mb.local is a self-hosted, fire-and-forget video compressor. Drop a file, choose a target size (e.g., 8 MB, 25 MB, 50 MB, 100 MB), and let GPU-accelerated encoding produce compact outputs with AV1/HEVC/H.264. Supports NVIDIA NVENC hardware encoding, Apple VideoToolbox, and automatic CPU fallback.

---

✨ Features

• NVIDIA NVENC hardware encoding natively supported via Container Toolkit
• Native macOS VideoToolbox hardware encoding via a transparent host daemon running securely via launchd for Apple M-series chips
• Two-Mode Interface: A streamlined Simple Mode and an advanced Sidebar Mode
• Smart Quality Profiles (Fast, Balanced, Quality, Best) with a Perfect Quality Override function
• Robust encoder validation at startup — tests actual encoder initialization, not just availability
• AV1, HEVC (H.265), and H.264 encoding natively and via deep ARM container optimizations (-DNATIVE=ON), including libsvtav1 and libaom-av1
• Ultra-lightweight Frontend: Native CSS SvelteKit implementation yielding instantaneous Node compilation.
• Batch processing — compress multiple files in a single operation
• Job history and Auto-download workflows included

📸 Screenshots

Home	Advanced	Batch
Queue	History	Settings - System Options
Settings - System Options	Settings - App Management	Settings - Defaults & Options
Settings - Defaults & Options	Settings - Security	API

🚀 Installation Guides

Default (NVIDIA GPU)

Requires NVIDIA Container Toolkit and a working docker run --rm --gpus all nvidia/cuda:12.2.0-base-ubuntu22.04 nvidia-smi on the host.

docker run -d \
  --name 8mb.local \
  --gpus all \
  -e NVIDIA_DRIVER_CAPABILITIES=compute,video,utility \
  -p 8001:8001 \
  -v ./data:/app/data \
  -v ./uploads:/app/uploads \
  -v ./outputs:/app/outputs \
  xesurient/8mb.local:latest

macOS Hardware Acceleration (M-Series)

Docker on macOS runs inside a Linux VM and fundamentally lacks access to Apple Silicon hardware encoders. 8mb.local ships with a native, zero-copy macOS daemon that intercepts encode requests using standard macOS launchctl bounds:

1. Bring up your stack using the CPU-only compose file: docker compose -f docker-compose.cpu.yml up -d --build
2. Install the native daemon on your host Mac:

   cd daemon
   bash install.sh

   or

   ./daemon/install.sh

The daemon runs natively on macOS. The Docker container will automatically bridge with your M-series hardware yielding huge encoding performance boosts.

Note: You can manually customize your VideoToolbox connection logic natively inside the 8mb.local instance Settings UI if you choose to bind your daemon to a custom port.

CPU Only (No GPU)

docker run -d \
  --name 8mb.local \
  -p 8001:8001 \
  -v ./data:/app/data \
  -v ./uploads:/app/uploads \
  -v ./outputs:/app/outputs \
  xesurient/8mb.local:latest

Access the web UI at http://localhost:8001. If building from source, simply clone the repository and run docker compose up -d --build. Use docker-compose.cpu.yml for CPU-only nodes.

💻 Automation & Headless API

8mb.local maintains an entirely distinct backend logic schema parsed natively through FastAPI. The web UX is completely optional.

FastAPI automatically generates an interactive swagger interface to test payloads. Assuming 8mb.local is active, simply navigate your web browser to http://localhost:8001/docs to view and construct valid curl configurations.

For shell-script automation and headless workflow structures, review the API Reference Guide.

📝 Credits

8mb.local utilizes elements from the broader open-source community. Special thanks to FFmpeg for the incredibly powerful encoding logic powering the backend, and to the Svelte framework for rendering the frontend footprint at a fraction of the cost of heavy JS frameworks.

UI assets and components were modeled and sourced using open design palettes, adopting clean native CSS practices for maximum speed. The initial homepage architecture is a fork and rebase of the beautiful fits.video project created by their open-source GitHub team, tweaked and structurally unified to cleanly serve the 8mb.local compressor workflow.

📚 Advanced Documentation

Need to configure something deep inside the stack? We've abstracted advanced operations out into targeted guides.

• ⚙️ Configuration & Environment Limits
• ⚡ Architecture & Concurrency Tuning
• 🛡️ Reverse Proxy SSE Routing (NGINX/Traefik)
• 🖥️ Supported Hardware Processors
• 📜 Internal Task Queue System