OpenTryOn

OpenTryOn is an open-source AI toolkit from TryOn Labs for fashion technology and virtual try-on. It puts virtual try-on, image and video generation and editing, multimodal understanding, background removal, preprocessing and datasets behind a single model registry. The latest release listed is v0.0.8, published on PyPI and GitHub.

What It Is

OpenTryOn is a developer toolkit that wraps many third-party and local fashion-AI models behind one consistent interface. Users supply inputs such as a person image and a garment image, pick a model, and get generated try-on images, edited images, videos, text descriptions, body measurements, embeddings or decisions. It also includes garment and pose preprocessing, fashion datasets (Fashion-MNIST, VITON-HD, Subjects200K), Gradio demos and TryOnDiffusion research code.

Four Ways to Run It

The same registry is exposed through four surfaces:

• A CLI of the form opentryon <service> --model <model>, with a --dry-run flag to verify wiring without calling an API.
• A FastMCP server that exposes every registry model as a tool for Claude, Cursor and other MCP clients.
• TryOn Studio, a Next.js playground with agent chat, a connect screen and capability screens for image, VTON, understand, video and background removal, talking to the MCP server over HTTP.
• Python adapters under tryon.api plus a single invoke_model() entry point.

Model and Service Coverage

Services include virtual try-on (vton), text-to-image (generate), edit, understand, video-generate, audio-to-video, generate-3d, bg-remove, tts, decide, classify, embed and measure. Try-on adapters listed include FLUX VTO, Google Vertex Virtual Try-On, FASHN, Photoroom, Segmind and Kling AI, along with local-weight models such as Leffa and CatVTON. Local and GPU models need the optional opentryon[local] install.

Setup Path

Install from PyPI with pip install opentryon or clone the repository and install into a conda environment, then copy env.template to .env and add the API keys for the providers you use. The docs list Python 3.10 or higher as a prerequisite and recommend a CUDA-capable GPU for best performance with local models.

Licensing

The project is released under Creative Commons BY-NC 4.0, which permits non-commercial use with attribution. The license file gives contact@tryonlabs.ai for commercial licensing inquiries.