Alchemist

Intelligent video transcoding automation with hardware acceleration and CPU fallback

Alchemist is a Rust-based video transcoding system that automatically converts your media library to efficient AV1 format using hardware acceleration (GPU) or software encoding (CPU fallback).

Features

• ** Hardware Acceleration**: Supports NVIDIA (NVENC), Intel (QSV), AMD (VAAPI), and Apple (VideoToolbox)
• ** CPU Fallback**: Automatic software encoding with libsvtav1 when GPU is unavailable
• ** Intelligent Analysis**: Only transcodes files that will benefit from AV1 encoding
• ** Web Dashboard**: Real-time monitoring and control via Leptos-based UI
• ** Background Processing**: Queue-based system with concurrent job support
• ** Performance Metrics**: Detailed logging and statistics for each transcode job

Quick Start

Prerequisites

• Rust: 1.75+ (nightly for WASM support)
• FFmpeg: With hardware acceleration support and libsvtav1
• FFprobe: For media analysis
• Docker (optional): For containerized deployment

Installation

# Clone the repository
git clone https://github.com/yourusername/alchemist.git
cd alchemist

# Build the project
cargo build --release

# Run the server
./target/release/alchemist --server

First-Boot Configuration

When you start Alchemist server for the first time, an interactive configuration wizard will run automatically:

# First time running the server
./target/release/alchemist --server

# The wizard runs automatically and guides you through:
# - Transcoding settings (quality thresholds, concurrency)
# - Hardware configuration (GPU/CPU encoding preferences)
# - Auto-scan directories (media library paths)
#
# After completion, config.toml is created and the server starts

On subsequent runs, the server loads your saved configuration.

Docker Deployment

# Build the Docker image
docker build -t alchemist .

# Run the container
docker run -d \
  -p 3000:3000 \
  -v /path/to/media:/media \
  -v /path/to/output:/output \
  --name alchemist \
  alchemist

Access the web interface at http://localhost:3000

Architecture

Components

┌─────────────────────────────────────────────────────────────┐
│                      Web Interface (Leptos)                 │
│                    http://localhost:3000                    │
└───────────────────────────┬─────────────────────────────────┘
                            │
┌───────────────────────────┴─────────────────────────────────┐
│                      Axum HTTP Server                       │
│                    (API + Static Assets)                    │
└───────────────────────────┬─────────────────────────────────┘
                            │
            ┌───────────────┼───────────────┐
            │               │               │
┌───────────▼────────┐ ┌────▼───────┐ ┌─────▼──────┐
│     Scanner        │ │  Analyzer  │ │  Database  │
│  (Media Discovery) │ │  (FFprobe) │ │  (SQLite)  │
└────────────────────┘ └────────────┘ └─────┬──────┘
                                            │
            ┌───────────────────────────────┘
            │
┌───────────▼────────────────────────────────────────┐
│              Processor (Job Queue)                 │
│          • Concurrent job execution                │
│          • State management                        │
└───────────┬────────────────────────────────────────┘
            │
┌───────────▼────────────────────────────────────────┐
│           Orchestrator (FFmpeg Wrapper)            │
│                                                    │
│  GPU Mode:           CPU Mode:                     │
│  • av1_qsv  (Intel)  • libsvtav1 (software)        │
│  • av1_nvenc (NVIDIA)                              │
│  • av1_vaapi (AMD)                                 │
└────────────────────────────────────────────────────┘

Database Location

Current Behavior:

• The SQLite database (alchemist.db) is created in the current working directory when the application starts
• This is intentional for flexibility during development and deployment

Future Configuration:

• A configurable database path will be added via:
  • Environment variable: ALCHEMIST_DB_PATH
  • Configuration file: config.toml
  • Docker mount point: /app/data/alchemist.db

Docker Recommendation: For production use with Docker, mount a persistent volume:

docker run -v /host/data:/app/data alchemist

Configuration

Create a config.toml file in the working directory:

[transcode]
size_reduction_threshold = 0.3  # Require 30% size reduction
min_bpp_threshold = 0.1          # Minimum bits per pixel
min_file_size_mb = 50            # Skip files smaller than 50MB
concurrent_jobs = 2              # Number of parallel transcodes

[hardware]
allow_cpu_fallback = true        # Enable CPU encoding if no GPU
allow_cpu_encoding = true        # Allow software encoding
cpu_preset = "medium"            # CPU encoding speed: slow|medium|fast|faster
# preferred_vendor = "intel"     # Force specific GPU vendor

[scanner]
directories = [                  # Auto-scan directories
    "/media/movies",
    "/media/tvshows"
]

Usage

CLI Mode

# Scan and transcode specific directories
alchemist /path/to/videos /another/path

# Dry run (analyze only, don't transcode)
alchemist --dry-run /path/to/videos

# Specify output directory
alchemist --output-dir /path/to/output /path/to/input

Server Mode

# Start web server on default port (3000)
alchemist --server

# Access web interface
open http://localhost:3000

Hardware Acceleration

Supported Platforms

Vendor	Encoder	Quality	Speed	Linux	macOS	Windows
Intel QSV	av1_qsv	Good	Fast	✅	✅	✅
NVIDIA NVENC	av1_nvenc	Excellent	Very Fast	✅	❌	✅
AMD VAAPI	av1_vaapi	Good	Fast	✅	❌	❌
Apple VideoToolbox	av1_videotoolbox	Excellent	Fast	❌	✅	❌
CPU (libsvtav1)	libsvtav1	Excellent	Slow	✅	✅	✅

CPU Encoding Performance

CPU encoding is 10-50x slower than GPU acceleration, depending on resolution:

Resolution	GPU Time	CPU Time (Preset 8)	Recommended
1080p	2-5 min	20-60 min	✅ Acceptable
4K	5-15 min	1-4 hours	⚠️ Slow
8K	15-45 min	4-12 hours	❌ Very Slow

CPU Preset Guide:

• slow (0-4): Best quality, extremely slow
• medium (5-8): Balanced (recommended for CPU)
• fast (9-13): Fastest, lower quality

Development

Project Structure

alchemist/
├── src/
│   ├── main.rs           # Entry point and CLI
│   ├── lib.rs            # Library exports
│   ├── server.rs         # Axum web server
│   ├── app.rs            # Leptos UI components
│   ├── hardware.rs       # GPU/CPU detection
│   ├── config.rs         # Configuration management
│   ├── scanner.rs        # Media file discovery
│   ├── analyzer.rs       # FFprobe wrapper
│   ├── processor.rs      # Job queue processor
│   ├── orchestrator.rs   # FFmpeg orchestration
│   ├── db.rs             # SQLite database
│   └── error.rs          # Error types
├── style/                # Tailwind CSS
├── public/               # Static assets
├── Cargo.toml            # Rust dependencies
├── Dockerfile            # Container definition
└── config.toml           # Configuration (optional)

Building for Development

# Install cargo-leptos for dev server with hot reload
cargo install cargo-leptos

# Run development server
cargo leptos watch

# Run tests
cargo test

# Check code
cargo check

Troubleshooting

GPU Not Detected

1. Check FFmpeg support:

   ffmpeg -encoders | grep av1
   

2. Verify device nodes (Linux):

   ls -la /dev/dri/render*
   

3. Check NVIDIA drivers:

   nvidia-smi
   

CPU Encoding Too Slow

1. Adjust preset in config.toml:

   [hardware]
   cpu_preset = "faster"  # Fastest CPU encoding
   

2. Reduce concurrent jobs:

   [transcode]
   concurrent_jobs = 1
   

License

This project is licensed under the MIT License - see the LICENSE file for details.