VectorSpaceLab/OmniGen2
OmniGen2 is an advanced multimodal image generation model supporting text-to-image, image-to-image, and instruction-guided image editing.
Not currently ranked — collecting fresh signals.
star history
OmniGen2 is a research project focused on multimodal generation, allowing users to generate and edit images from text instructions and other images. The project includes model weights, a Gradio demo, training/inference code, and benchmarks like EditReward-Bench for evaluating image editing quality. It also releases reward models (EditScore family, 7B–72B) for reinforcement learning-based image editing.
Frequently asked
- What is VectorSpaceLab/OmniGen2?
- OmniGen2 is an advanced multimodal image generation model supporting text-to-image, image-to-image, and instruction-guided image editing.
- Is OmniGen2 open source?
- Yes — VectorSpaceLab/OmniGen2 is open source, released under the Apache-2.0 license.
- What language is OmniGen2 written in?
- VectorSpaceLab/OmniGen2 is primarily written in Jupyter Notebook.
- How popular is OmniGen2?
- VectorSpaceLab/OmniGen2 has 4.1k stars on GitHub.
- Where can I find OmniGen2?
- VectorSpaceLab/OmniGen2 is on GitHub at https://github.com/VectorSpaceLab/OmniGen2.