tryonlabs/opentryon

Open-source APIs, SDKs, and models for building virtual try-on and fashion AI applications. Generate models, edit garments, create photoshoots, and build personalized fashion experiences.

OpenTryOn (opentryon) – Quick Overview

What it is – An open‑source Python toolkit that bundles a large collection of fashion‑focused AI models and services. It lets developers run virtual‑try‑on, image/video generation, multimodal understanding, background removal, and related preprocessing either locally or via cloud APIs, all through a unified command‑line interface (CLI), a Python library, or a web UI called TryOn Studio.

Core Capabilities

Area What you can do Example models (as listed in the repo)
Virtual Try‑On (VTO) Swap garments onto a person image, generate a model wearing a product, re‑frame outfits, convert images to video. flux-vto, google-vertex-vto, outfitanyone-plus, photoroom, fashn-tryon-max, klein-4b (commercially clean), catvton
Image Generation / Editing Text‑to‑image, style transfer, up‑scaling, composition. nano-banana-pro, flux2-pro, gpt-image-2.5, ideogram-4.0, pruna-p-image, qwen-image
Multimodal Understanding Describe images/video, extract outfit metadata, answer questions about fashion content. claude-haiku-5.5, kimi-k3, qwen3.8-max, nemotron‑omni, hy4-preview
Video Generation Turn text or images into short fashion videos, animate outfits. veo-3.1, luma-ray-2, seedance-2.5, ltx-2.5, runway‑gen‑4.5, nvidia‑cosmos‑3
Background Removal & Pre‑processing Cleanly cut garments or people from photos, prepare data for training. ben2
Safety & Decision Tools Classify whether a try‑on request is safe, make calibrated yes/no decisions. tryon-safety-classifier, jev, d1-3b
Embeddings Produce multimodal feature vectors for retrieval or downstream models. embeddinggemma-2, pplx-embed-v2

How to Use It

  1. CLI – One‑liner commands, e.g.:
    opentryon vton --model flux-vto \
  --person-image data/model-1.jpg \
  --garment-image data/garment.png
  2. MCP Server – Starts a FastMCP server that exposes every model as a tool for agents like Claude, Cursor, or the TryOn Studio UI.
  3. TryOn Studio – A Next.js web interface that calls the MCP server under the hood, offering a point‑and‑click experience for VTO, generation, understanding, etc.
  4. Python Library – Import adapters directly:
    from tryon.api import KimiUnderstandAdapter
adapter = KimiUnderstandAdapter(model="kimi-k3")
result = adapter.understand_image("data/model-1.jpg", prompt="Describe this outfit.")
print(result["text"])

Installation Snapshot

# Clone & set up conda environment (recommended)
git clone https://github.com/tryonlabs/opentryon.git
cd opentryon
conda env create -f environment.yml
conda activate opentryon
pip install -e .               # core package
pip install -e "[local]"      # optional local/GPU models
cp env.template .env          # fill in required API keys (e.g., BFL, Moonshot)

Or simply:

pip install opentryon

Project Layout (high‑level)

  • tryon/ – Core package (API wrappers, CLI, model registry, agents, datasets, preprocessing).
  • tryondiffusion/ – Research code for diffusion‑based try‑on models.
  • mcp-server/ – FastMCP server implementation.
  • demo/ – Gradio demo scripts.
  • notebooks/ – Jupyter examples.
  • docs/ – Docusaurus site (full docs, tutorials, road‑map).
  • openapi/ & postman/ – Swagger / Postman collections for external media APIs.

Documentation & Community

  • Full docs – https://tryonlabs.github.io/opentryon/ (installation, CLI guide, MCP setup, agent usage, model‑specific pages).
  • Discord – Community chat for support and feature requests.
  • Roadmap – Publicly listed in ROADMAP.md.
  • License – Creative Commons BY‑NC 4.0 (non‑commercial use with attribution).

Who Might Use This?

  • Fashion e‑commerce platforms wanting to add virtual try‑on without building everything from scratch.
  • Developers & researchers experimenting with diffusion or multimodal models for clothing.
  • Product teams building internal tools (e.g., catalog generation, outfit description, background removal).
  • AI agents that need fashion‑aware tools exposed via MCP.

Bottom Line

OpenTryOn is a comprehensive, extensible toolkit that unifies dozens of cutting‑edge fashion‑related AI models behind a single, easy‑to‑use interface. It supports both cloud API calls and local GPU inference, making it suitable for rapid prototyping as well as production‑grade deployments in the fashion tech space.

Related

  • Project
    aimagelab/multimodal-garment-designer

    A multimodal fashion image editing tool based on latent diffusion models that generates human-centric images from text, body poses, and garment sketches.

  • Project
    voxel51/fiftyone

    An open-source tool for building high-quality computer vision datasets and models by providing visual data exploration, labeling, and model evaluation.

  • Project
    AIFrontierLab/TorchUMM

    A unified framework for the inference, evaluation, and post-training of multimodal models, supporting image understanding, generation, and editing through a single interface.

  • Dispatch
    OpenAI announces partnership with PVH on future of fashion

    OpenAI announced a collaboration with PVH to explore AI-driven innovations in fashion, highlighting the strategic importance of AI for the apparel industry.

  • Project
    triton-inference-server/server

    An open-source inference serving software that streamlines the deployment of AI models from multiple frameworks across cloud, data center, edge, and embedded devices.