inclusionAI/Ming-UniVision
Code release for Ming-UniVision: Joint Image Understanding and Geneation with a Continuous Unified Tokenizer
This project helps creative professionals and researchers who work with both images and text by providing a way to seamlessly generate, understand, and edit visual content using natural language. You can input text prompts or existing images, and the system can produce new images, answer questions about images, or modify parts of an image through conversational interaction. It's for designers, content creators, or AI researchers exploring advanced multimodal capabilities.
142 stars. No commits in the last 6 months.
Use this if you need to create, interpret, or alter images in a continuous, conversational flow without converting between different formats or states.
Not ideal if your primary need is simple image classification, object detection, or if you require extremely fine-grained, pixel-level control not driven by high-level semantic instructions.
Stars
142
Forks
5
Language
Python
License
MIT
Category
Last pushed
Oct 14, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/inclusionAI/Ming-UniVision"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
Vchitect/VBench
[CVPR2024 Highlight] VBench - We Evaluate Video Generation
VectorSpaceLab/OmniGen
OmniGen: Unified Image Generation. https://arxiv.org/pdf/2409.11340
EndlessSora/focal-frequency-loss
[ICCV 2021] Focal Frequency Loss for Image Reconstruction and Synthesis
JIA-Lab-research/DreamOmni2
This project is the official implementation of 'DreamOmni2: Multimodal Instruction-based Editing...
SkyworkAI/UniPic
Open-source SOTA multi-image editing model