Skip to content

Releases: tryonlabs/opentryon

🚀 OpenAI Integration & AI Agents - Enhanced Capabilities for Fashion AI

27 Dec 07:12
@kailashahirwar kailashahirwar
v0.0.2
32c61aa

Choose a tag to compare

View all tags
🚀 OpenAI Integration & AI Agents - Enhanced Capabilities for Fashion AI Latest
Latest

OpenTryOn v0.0.2 Release Notes

Release Date: 27 December 2025

We're excited to announce the release of OpenTryOn v0.0.2! This release introduces significant new features including OpenAI integrations for image and video generation, Google Veo 3 video generation, and two powerful AI agents for virtual try-on and model swapping.

🎉 What's New

OpenAI Integration

Image Generation - GPT-Image Models

We've added full support for OpenAI's GPT-Image models, bringing professional-grade image generation capabilities to OpenTryOn:

  • GPT-Image-1.5 (Default): The latest model with enhanced quality, better prompt understanding, and improved consistency
  • GPT-Image-1: High-quality image generation and editing

Features:

  • Text-to-image generation
  • Image-to-image editing
  • Multi-image composition
  • Mask-based editing with precise region control
  • Background control and quality settings
  • Multiple output images support

Example Usage:

from tryon.api.openAI import GPTImageAdapter

# Initialize adapter (defaults to GPT-Image-1.5)
adapter = GPTImageAdapter()

# Generate image from text
images = adapter.generate_text_to_image(
    prompt="A fashion model wearing an elegant evening gown",
    size="1024x1024",
    quality="high"
)

# Edit existing image
images = adapter.generate_image_edit(
    images="person.jpg",
    prompt="Change the dress color to midnight blue"
)

Documentation: GPT-Image API Reference

Video Generation - Sora Models

OpenTryOn now supports OpenAI's Sora video generation models for creating high-quality videos:

  • Sora 2 (Default): Fast, high-quality video generation
  • Sora 2 Pro: Enhanced quality with superior temporal consistency

Features:

  • Text-to-video generation
  • Image-to-video generation (animate static images)
  • Support for 4, 8, and 12-second videos
  • Multiple resolutions (720p to Full HD)
  • Two wait modes: synchronous (blocking) and asynchronous (callback-based)
  • Progress tracking and status monitoring

Example Usage:

from tryon.api.openAI import SoraVideoAdapter

# Initialize adapter
adapter = SoraVideoAdapter()

# Generate video from text
video_bytes = adapter.generate_text_to_video(
    prompt="A fashion model walking on a runway, elegant movements",
    duration=8,
    resolution="1280x768"
)

# Animate an image
video_bytes = adapter.generate_image_to_video(
    image="model.jpg",
    prompt="The model gracefully walking",
    duration=12
)

Documentation: Sora Video API Reference

Google Veo 3 Video Generation

We've integrated Google's Veo 3 video generation model, providing another powerful option for creating cinematic videos:

Features:

  • High-quality, cinematic video generation
  • Text-to-video and image-to-video support
  • Realistic motion and temporal consistency
  • Fine-grained control over style and camera dynamics

Example Usage:

from tryon.api import VeoAdapter

adapter = VeoAdapter()
video_bytes = adapter.generate_text_to_video(
    prompt="A fashion show with models showcasing elegant designs"
)

Documentation: Veo Video Documentation

AI Agents

Virtual Try-On Agent (VTOnAgent)

A new intelligent agent that automates virtual try-on operations using LangChain:

Features:

  • Automatically analyzes prompts and selects appropriate models
  • Support for multiple virtual try-on providers (Kling AI, Segmind, Nova Canvas)
  • Tool-based architecture for flexible model selection
  • Intelligent decision-making for optimal results

Example Usage:

from tryon.agents import VTOnAgent

# Initialize the agent
agent = VTOnAgent(llm_provider="openai")

# Generate virtual try-on
result = agent.generate(
    person_image="person.jpg",
    garment_image="shirt.jpg",
    prompt="Use Kling AI to create a virtual try-on of this shirt"
)

Documentation: VTOn Agent Documentation

Model Swap Agent (ModelSwapAgent)

An AI agent that replaces models in images while preserving outfit consistency:

Features:

  • Automatically swaps models while maintaining outfit consistency
  • Support for multiple image generation models (Nano Banana, Nano Banana Pro, FLUX 2 Pro, FLUX 2 Flex)
  • Intelligent prompt engineering for model swapping
  • Preserves outfit details and styling

Example Usage:

from tryon.agents import ModelSwapAgent

# Initialize agent with default Nano Banana Pro
agent = ModelSwapAgent(llm_provider="openai")

# Generate model swap
result = agent.generate(
    image="person_wearing_outfit.jpg",
    prompt="Replace with a professional Asian female model in her 30s, athletic build"
)

Documentation: Model Swap Agent Documentation

📦 Installation

Upgrade to v0.0.2 using pip:

pip install --upgrade opentryon

Or install from source:

git clone https://github.com/tryonlabs/opentryon.git
cd opentryon
git checkout v0.0.2
pip install -e .

🔧 Configuration

OpenAI API Key

For GPT-Image and Sora features, set your OpenAI API key:

export OPENAI_API_KEY="your-openai-api-key"

Or add to your .env file:

OPENAI_API_KEY=your-openai-api-key

Google Veo API Key

For Veo video generation, set your Gemini API key:

export GEMINI_API_KEY="your-gemini-api-key"

📚 Documentation Updates

  • Added comprehensive API reference for GPT-Image models
  • Added comprehensive API reference for Sora video models
  • Added documentation for VTOn Agent
  • Added documentation for Model Swap Agent
  • Added Veo 3 video generation documentation
  • Updated quickstart guides with new features
  • Enhanced examples and use cases

🔄 Migration Guide

From v0.0.1 to v0.0.2

No breaking changes! All existing code should continue to work. The new features are additive:

  • Existing virtual try-on adapters remain unchanged
  • Existing image generation adapters remain unchanged
  • New OpenAI and agent modules are additional features
  • No changes required to existing code

🐛 Bug Fixes

  • Fixed broken link in fashion-prompt-builder documentation

🙏 Acknowledgments

Thank you to all contributors who made this release possible! Special thanks to:

  • Contributors who added OpenAI integration
  • Contributors who implemented the AI agents
  • Community members who provided feedback and testing

📝 Full Changelog

For a complete list of changes, see CHANGELOG.md.

🔗 Links

📄 License

This release is licensed under CC BY-NC 4.0. See LICENSE for details.

Note: This is an alpha release. We welcome feedback and contributions from the community!

Assets 2
Loading

🎉 OpenTryOn v0.0.1 - First Public Release

16 Dec 14:11
@kailashahirwar kailashahirwar
v0.0.1
df9881c

Choose a tag to compare

View all tags
🎉 OpenTryOn v0.0.1 - First Public Release

🚀 What's New

We're thrilled to announce the first public release of OpenTryOn - an open-source AI toolkit for fashion tech and virtual try-on applications!

OpenTryOn is now available on PyPI:

pip install opentryon

✨ Features

Virtual Try-On (3 Providers)

  • Amazon Nova Canvas - AWS Bedrock-powered virtual try-on
  • Kling AI - Kolors-based high-quality try-on
  • Segmind - Fast Try-On Diffusion API

Image Generation (6 Models)

  • Gemini 2.5 Flash & 3 Pro - Fast 1024px to 4K generation with search grounding
  • FLUX.2 PRO & FLEX - High-quality generation with advanced controls
  • Luma AI Photon-1 & Photon-Flash-1 - Professional-grade image generation

Video Generation

  • Luma AI Ray Models - Text-to-video and image-to-video (Ray 1.6, Ray 2, Ray Flash 2)

Datasets

  • Fashion-MNIST - Automatic download and loading
  • VITON-HD - High-resolution virtual try-on dataset with lazy loading

Preprocessing Tools

  • Garment segmentation (U2Net)
  • Garment extraction and preprocessing
  • Human segmentation and parsing

📦 Installation

Quick Start

pip install opentryon

From Source

git clone https://github.com/tryonlabs/opentryon.git
cd opentryon
pip install -e .

Configuration

# Copy environment template
cp env.template .env

# Add your API keys to .env
# Minimum required: One virtual try-on + one image generation service

🎯 Quick Example

from tryon.api.kling_ai import KlingAIVTONAdapter

# Virtual try-on
adapter = KlingAIVTONAdapter()
result = adapter.virtual_tryon(
    model_image="person.jpg",
    garment_image="shirt.jpg"
)
result.save("output.jpg")

📚 Documentation

  • Full Documentation: https://tryonlabs.github.io/opentryon/
  • API Reference: Complete API docs for all modules
  • Examples: Usage examples for all features
  • Getting Started: Step-by-step tutorials

🔧 Requirements

  • Python 3.10+
  • At least one API key (Kling AI, Gemini, FLUX.2, etc.)
  • GPU recommended for optimal performance

📋 What's Included

  • ✅ 3 virtual try-on providers
  • ✅ 6 image generation models
  • ✅ Video generation capabilities
  • ✅ Dataset loaders (Fashion-MNIST, VITON-HD)
  • ✅ Complete preprocessing pipeline
  • ✅ Interactive Gradio demos
  • ✅ Comprehensive documentation

🎨 Use Cases

  • E-commerce: Virtual try-on for online shopping
  • Fashion Brands: Automated product photography
  • Content Creation: Fashion imagery for marketing
  • AI Agents: Fashion tech capabilities for agents
  • Research: Training and benchmarking

🚨 Known Limitations

  • Amazon Nova Canvas requires AWS account (optional)
  • Luma AI required only for video generation (optional)
  • U2Net preprocessing requires checkpoint download (optional)
  • Some features require GPU for optimal performance

🔐 License

Creative Commons BY-NC 4.0 (Non-commercial use)

For commercial licensing, contact: contact@tryonlabs.ai

🙏 Acknowledgments

Special thanks to:

  • The open-source AI community
  • API providers: AWS, Google, Black Forest Labs, Luma AI, Kling AI, Segmind
  • All contributors and early testers

🔗 Resources

📈 What's Next

  • OminiControl2 training support
  • Additional virtual try-on models
  • Enhanced preprocessing tools
  • More dataset loaders
  • Community-driven features

💬 Get Involved

  • ⭐ Star the repository
  • 🍴 Fork and contribute
  • 🐛 Report issues
  • 💡 Suggest features
  • 💬 Join our Discord

Made with ❤️ by TryOn Labs

Released: December 16, 2025

Loading