The news of January 28thAlibaba CloudGeneral meaningOfficial today, 28 JanuaryOpen Sourcerelease Z-Image Base base model.
THE MODEL SIZE IS 6B, maintain full weight distribution for the non-distillation base model, original support for the CFG pilot mechanism, and provide a training base for fine-tuning missions such as LoRA, ContractorNet。
Z-Image claims to break the writing limits of a single dimension: Whether it is phototolealism that pursues light, or the dynamic and digital arts that have emotional tension, Z-Image captures and reconstructs the details of each style。
Z-Image uses a non-distillation base structure, preserves the full weight distribution, supports the CFG pilot mechanism, and provides a training base for fine-tuning missions such as LoRA, ContractorNet。
In addition, the Z-Image special purpose optimizes the generation of “homogenization” pain points: optimizes the spatial distribution of sampling and ensures that the faces and structures of different originals differ significantlyMULTIPLE SCENES STRIPPING OFF INDIVIDUAL FEATURES AND REJECTING AI "PEOPLE'S FACE".
Z-Image can also respond sensitively to negative hints and achieve Depth control from the image to the light through the Negative Prompt rapid filtering of defective and purified images。
1AI with Z-Image open source addresses as follows:
- GitHub: https://github.com/Tongyi-MI/Z-Image
- Magic: https://modersscope.cn/models/Tongyi-MI/Z-Image
- Hugging Face: https://huggingface.co/Tongyi-MI/Z-Image