Local Deployment Guide
Run the Happy Horse 1.0 (15B Unified Transformer) on your own hardware for maximum control and privacy.
Hardware Requirements
GPU Recommendation
NVIDIA H100 or A100
Minimum 48GB VRAM for 1080p output.
System RAM
64GB+ DDR5
Necessary for model loading and processing.
Storage
100GB+ NVMe SSD
For model weights and cache.
Performance
~38s per 5s clip
Benchmarked on NVIDIA H100 at 1080p.
bash — Quick Start
# Clone the repository
git clone https://github.com/HappyHorse/HappyHorse-1.0.git
# Install dependencies
pip install -r requirements.txt
# Launch the inference server
python main.py --model 15B --device cuda --precision bf16
Commercial Use
Happy Horse 1.0 weights are released under a commercial-use friendly license. You are free to use generated videos for production and social media monetization.