OmniGen2: Unified Multimodal Generation Model by VectorSpace Lab

OmniGen2 is a powerful unified multimodal generation model supporting text-to-image, instruction-based image editing, in-context generation, and visual understanding.

O

OmniGen2

Text-to-ImageImage EditingIn-Context GenerationVisual Understanding

Unified multimodal generation model by VectorSpace Lab. Combines text-to-image generation, instruction-guided image editing, in-context generation, and visual understanding in a single model. Uses decoupled decoding pathways for text and image modalities, built on a Qwen-VL-2.5 foundation.

Hugging FaceHF Comfy-OrgGitHubPaper
DeveloperVectorSpace Lab
Release Date2025-06-16
ArchitectureUnified Multimodal with Decoupled Decoding Pathways
LicenseApache-2.0
Text EncoderQwen2.5-VL (FP16)
CapabilitiesT2I, image editing, in-context gen, visual understanding

Guides and workflows related to this model series.

No articles found.