Welcome to the experimental phase of Z Image MEeE, a versatile checkpoint built upon the powerful Z-Image base. This model is meticulously trained to strike a perfect balance between lifelike human subjects, intricate objects, and breathtaking scenery.

Whether you are generating sweeping landscapes, detailed everyday objects, or realistic portraits, Z Image MEeE aims to deliver depth, rich textures, and a strong sense of realism.

• Base Model: Z-Image Base
• Training Core: ~6,000 Curated Images / 14,000 Regularization Images
• Primary Focus: Photorealism, Scenery, General Objects, and Humans.

⚠️ Important Disclaimer & Warnings

Please read before using this model:

• Work in Progress: This model is currently in a highly experimental stage. Because it is still being actively developed and fine-tuned, you may occasionally experience inconsistent, unoptimized, or less-than-ideal results.
• Content Potential: Due to the nature of the training dataset and the flexibility of the base model, this checkpoint can generate content. Please use appropriate prompt filtering (or negative prompts) if you wish to keep your generations strictly SFW.

Recommended Settings

To get the most out of this experimental version, we suggest starting with these baseline parameters:

• Sampler: DPM++ 2M Karras or Euler a
• Sampling Steps: 25 – 35
• CFG Scale: 4.5 – 7.0 (Lower values often yield softer, more natural realism)
• Resolution: Best optimized for 1024x1024 (1:1), 896x1152 (Portrait), or 1152x896 (Landscape).
• Prompting Style: Detailed descriptive prompts work best. Feel free to use photography terms (e.g., 8k resolution, cinematic lighting, photorealistic) to enhance the output.

Feedback Welcome: Since this is a test build, your generations and reviews are incredibly valuable! Please share your images and settings in the community gallery to help improve the next iteration.

V1. Epoch 73