OmniGen

Apps

VectorSpaceLab /

Extract specific characters from input images, and reimagine them in novel scenarios, using simple instructions.

Manipulate image elements using flexible descriptors such as position, age, appearance, and more.

Detect specific objects in input images and combine them into a new image via natural text descriptions.

Edit images to remove or change elements while maintaining consistency, all without inpainting or controlnet.

Simulate OpenPose preprocessing without using any additional networks or tools.

Description

A unified image generation model for multi-modal prompting. Instead of loading additional network modules, such as ControlNet, and performing extra processing steps, such as inpainting, OmniGen can generate and manipulate images directly using natural language.

Interface

Web UI

HTTP API

Source Code

VectorSpaceLab/
OmniGen

Default Version

0a59959

License

MIT License

Image Generation