config.json

{
  "model_type": "z-anime",
  "architecture": "S3-DiT",
  "parameters": "6B",
  "license": "apache-2.0",
  "base_model": "Tongyi-MAI/Z-Image",
  "base_model_relation": "finetune",
  "author": "mcuo",
  "pipeline_tag": "text-to-image",
  "prompting": {
    "style": "natural-language",
    "negative_prompt_support": {
      "base": "full",
      "distill_8step": "limited",
      "distill_4step": "limited"
    }
  },
  "variants": {
    "base": {
      "bf16": "diffusion_models/z-anime-base-bf16.safetensors",
      "fp8": "diffusion_models/z-anime-base-fp8.safetensors",
      "recommended_settings": {
        "steps": "28-50",
        "cfg": "3.0-5.0",
        "sampler": "euler_ancestral",
        "scheduler": "beta"
      }
    },
    "distill_8step": {
      "bf16": "diffusion_models/z-anime-distill-8step-bf16.safetensors",
      "fp8": "diffusion_models/z-anime-distill-8step-fp8.safetensors",
      "recommended_settings": {
        "steps": 8,
        "cfg": 1.0,
        "sampler": "euler_ancestral",
        "scheduler": "beta"
      }
    },
    "distill_4step": {
      "bf16": "diffusion_models/z-anime-distill-4step-bf16.safetensors",
      "fp8": "diffusion_models/z-anime-distill-4step-fp8.safetensors",
      "recommended_settings": {
        "steps": 4,
        "cfg": 1.0,
        "sampler": "euler_ancestral",
        "scheduler": "beta"
      }
    },
    "gguf": {
      "q8_0": {
        "file": "gguf/z-anime-base-q8_0.gguf",
        "description": "Q8_0 quantization",
        "size": "~6.73 GB"
      },
      "q4_k_s": {
        "file": "gguf/z-anime-base-q4_k_s.gguf",
        "description": "Q4_K_S quantization",
        "size": "~4.2 GB"
      }
    }
  },
  "diffusers_folder": {
    "path": "diffusers/",
    "pipeline_class": "ZImagePipeline",
    "usage": "ZImagePipeline.from_pretrained('mcuo/Anime-Z', subfolder='diffusers', torch_dtype=torch.bfloat16)",
    "components": [
      "model_index.json",
      "scheduler/",
      "tokenizer/",
      "text_encoder/",
      "transformer/",
      "vae/"
    ]
  },
  "components": {
    "text_encoders": {
      "default": {
        "bf16": "text_encoder/qwen_3_4b-bf16.safetensors",
        "fp8": "text_encoder/qwen_3_4b-fp8.safetensors",
        "description": "Standard Z-Image text encoder, repackaged as a single safetensors",
        "comfyui_path": "ComfyUI/models/clip/"
      },
      "engineer_v4": {
        "bf16": "text_encoder/qwen_3_4b-engineer-v4-bf16.safetensors",
        "fp8": "text_encoder/qwen_3_4b-engineer-v4-fp8.safetensors",
        "description": "Alternative full fine-tune by BennyDaBall (SMART training, more varied outputs)",
        "source": "https://huggingface.co/BennyDaBall/Qwen3-4b-Z-Image-Engineer-V4",
        "comfyui_path": "ComfyUI/models/clip/"
      }
    },
    "vae": {
      "file": "vae/ae.safetensors",
      "description": "Z-Image VAE (slightly trained alongside Z-Anime)",
      "comfyui_path": "ComfyUI/models/vae/"
    }
  },
  "comfyui_paths": {
    "diffusion_models": "ComfyUI/models/diffusion_models/",
    "unet": "ComfyUI/models/unet/",
    "clip": "ComfyUI/models/clip/",
    "vae": "ComfyUI/models/vae/",
    "checkpoints": "ComfyUI/models/checkpoints/"
  },
  "requirements": {
    "custom_nodes": [
      "rgthree-comfy",
      "ComfyUI-Lora-Manager",
      "ComfyUI-SeedVR2_VideoUpscaler (optional)"
    ]
  },
  "supported_vram": "8GB+",
  "links": {
    "civitai": "https://civitai.red/models/2483351",
    "base_model": "https://huggingface.co/Tongyi-MAI/Z-Image",
    "engineer_v4": "https://huggingface.co/BennyDaBall/Qwen3-4b-Z-Image-Engineer-V4",
    "author": "https://huggingface.co/mcuo"
  },
  "notes": [
    "BF16 and FP8 are the main release formats.",
    "GGUF variants are intended for lower-memory or alternative inference setups.",
    "Two text encoders are included: the standard Z-Image one (default) and BennyDaBall's Engineer V4 (alternative).",
    "The diffusers/ subfolder is a full diffusers-format checkpoint loadable via ZImagePipeline.from_pretrained(repo, subfolder='diffusers')."
  ]
}