BEYOND_REALITY_Z_IMAGE - a photorealistic people-focused model based on Z-Image Turbo
Contents
While looking at Z-Image derivative models, I noticed a checkpoint called BEYOND_REALITY_Z_IMAGE on ModelScope. It is a fine-tune specialized for photorealistic people, with a film-photography-like texture.
Model overview
| Item | Details |
|---|---|
| Author | Nurburgring (Zhang Chi) |
| Base | Z-Image Turbo |
| Method | LoRA training on a people dataset, then merged |
| License | Apache License 2.0 |
| Versions | v1.0 to v3.0 |
| Downloads | About 3,800 |
ModelScope: Nurburgring/BEYOND_REALITY_Z_IMAGE
Features
- Skin texture optimization: More detailed skin texture on people
- Film-photography aesthetics: Color and lighting feel closer to film cameras
- Environmental detail: Backgrounds and props are rendered more carefully
- Keeps Z-Image Turbo’s speed: Can generate in 10 to 15 steps
Recommended settings
| Setting | Value |
|---|---|
| Sampler | Euler + Simple |
| Steps | 10-15 |
| CFG | 1-2 |
Because it is based on Z-Image Turbo, you can still get high-quality results with fewer steps. A lower CFG is enough too.
Will it run on an M1 Max 64GB?
Conclusion: yes.
Z-Image Turbo memory requirements
| Component | Size |
|---|---|
| Z-Image Turbo (BF16) | About 12 GB |
| Qwen3 4B text encoder | About 7 GB |
| VAE | A few hundred MB |
| Total | About 20 GB |
An M1 Max with 64 GB of unified memory has plenty of room. A quantized version is even lighter.
| Quantization | Size |
|---|---|
| BF16 (full) | About 12 GB |
| FP8 | About 6 GB |
| Q4_K_M (GGUF) | About 4 GB |
Running on Apple Silicon
It works with ComfyUI + Metal support. If there is an MLX version, it could be optimized even further. There are also reports that stable-diffusion.cpp can run the GGUF-quantized version with only 4 GB of VRAM.
Where Z-Image derivatives fit
The Z-Image family is roughly organized like this:
| Model | Use |
|---|---|
| Z-Image | Base model. Supports LoRA and ControlNet |
| Z-Image-Turbo | Distilled version. Fast generation in 8 steps |
| Z-Image-Omni-Base | Multimodal base model |
| Z-Image-Edit | Image-editing specialization |
BEYOND_REALITY_Z_IMAGE is a checkpoint that merges a photorealistic people-focused LoRA on top of Z-Image Turbo. It keeps Turbo’s speed while improving the quality of portraits.
Related articles
Z-Image and image-generation articles on this blog:
- Z-Image - Alibaba’s image generation AI said to surpass FLUX
- Can Z-Image run on RunPod?
- Generative AI: I need LLM / RAG at work, so I tried making LoRA in my spare time 2025 (Part 1)
- Build a LoRA training setup on an RTX 3060 Laptop (6 GB VRAM)