gemneye/omnitry-runpod repository overview

⁠OmniTry RunPod Container

A containerized version of OmniTry⁠ optimized for RunPod deployment. OmniTry is an AI-powered virtual try-on application that allows users to try on clothes virtually using advanced diffusion models.

⁠🚀 Quick Start on RunPod

1. Create a new pod on RunPod
2. Select this container: gemneye/omnitry-runpod:latest
3. Configure settings:
   • Port: 7860 (Gradio web interface)
   • GPU: Recommended RTX 3090 or better
   • VRAM: Minimum 12GB
4. Start the pod and wait for setup to complete (first run takes 10-15 minutes)
5. Access the interface via the provided RunPod URL

⁠🛠 Container Architecture

This container uses a runtime setup approach optimized for RunPod:

• Base Image: nvidia/cuda:12.3.2-cudnn9-runtime-ubuntu22.04
• Build Time: Only base image installation
• Runtime: All dependencies installed when container starts

⁠Runtime Installation Process

When the container starts, it automatically:

1. ✅ Installs Miniconda and Python 3.11
2. ✅ Installs PyTorch 2.4.0 with CUDA 12.4 support
3. ✅ Clones the OmniTry repository
4. ✅ Downloads required models:
   • FLUX.1-Fill-dev (~12GB)
   • OmniTry unified model
   • OmniTry clothes model
5. ✅ Installs Python dependencies
6. ✅ Installs Flash Attention for performance
7. ✅ Starts Gradio web interface on port 7860

⁠📋 System Requirements

• GPU: NVIDIA GPU with 12GB+ VRAM (RTX 3090/4090 recommended)
• RAM: 16GB+ system RAM
• Storage: 25GB+ free space for models and cache
• Architecture: AMD64 only

⁠🔧 Environment Variables

Optional environment variables you can set in RunPod:

HF_TOKEN=your_huggingface_token  # If needed for private models
GRADIO_SERVER_NAME=0.0.0.0       # Server bind address (default)
GRADIO_SERVER_PORT=7860          # Server port (default)

⁠📊 Model Information

This container automatically downloads:

• FLUX.1-Fill-dev: Base diffusion model (~12GB)
• OmniTry Unified: Main virtual try-on model
• OmniTry Clothes: Specialized clothing model

Models are cached between runs when using persistent storage.

⁠🔍 Troubleshooting

⁠Container Won't Start

• Ensure sufficient VRAM (12GB+ required)
• Check RunPod logs for setup progress
• First run requires 10-15 minutes for model downloads

⁠Out of Memory Errors

• Use GPU with more VRAM
• Close other GPU applications
• Restart the pod

⁠Slow Performance

• Ensure CUDA drivers are properly installed
• Use recommended RTX 3090 or better
• Check GPU utilization in RunPod

⁠🔗 Links

• Original Project: Kunbyte-AI/OmniTry⁠
• RunPod Container: sruckh/OmniTry-Runpod⁠
• Docker Hub: gemneye/omnitry-runpod⁠

⁠📝 License

This container follows the same license as the original OmniTry project. Please refer to the original repository⁠ for licensing information.

Built for RunPod • AMD64 Architecture • CUDA 12.3 Ready