Z-Image Base: A Foundation for High-Quality Image Generation
Z-Image Base is the main model in the ⚡️-Image family. It was made to create high-quality images that are diverse, cover many styles, and stick closely to what you ask for.
Unlike its faster version, Z-Image-Turbo, Z-Image Base is a full, non-simplified transformer. It's meant to be a key tool for creators, researchers, and developers who want complete creative freedom.
Key Features of Z-Image Base
Complete Foundation: Since it's a full, non-simplified model, Z-Image Base keeps all its training information. This allows for full Classifier-Free Guidance (CFG), which gives you the accuracy needed for detailed prompts and professional work.
Flexible Aesthetics: Great at many visual styles, including super-realistic photos, cinematic digital art, complex anime, and various stylized illustrations. It's perfect for situations where you need rich, varied artistic expression.
More Diverse Outputs: Offers much more variety in composition, faces, and lighting with different seeds. This means scenes with multiple people will look distinct and dynamic.
Strong Negative Control: The model responds very accurately to negative prompts, helping users get rid of unwanted elements and improve their compositions.
Recommended Settings for Z-Image Base
Resolution: You can make images from 512x512 up to 2048x2048 (any aspect ratio is fine, it's about the total pixels).
Guidance Scale: A guidance(CFG) scale between 3.0 and 5.0 is suggested.
Inference Steps: Use 28 to 50 inference steps to generate images.
Prompt Style: Longer, more detailed prompts work best (just like with Z-Image Turbo).
