Upload config.json

config.json

@@ -10,24 +10,65 @@
   "prompting": {
     "style": "natural-language",
     "negative_prompt_support": {
-      "distill_8step": "limited"
+      "base": "full",
+      "distill_8step": "limited",
+      "distill_4step": "limited"
     }
   },
   "variants": {
+    "base": {
+      "bf16": "diffusion_models/z-anime-base-bf16.safetensors",
+      "fp8": "diffusion_models/z-anime-base-bf16.safetensors",
+      "aio_bf16": "diffusion_models/z-anime-base-bf16.safetensors",
+      "aio_fp8": "diffusion_models/z-anime-base-bf16.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",
+      "bf16": "diffusion_models/z-anime-distill-8step-fp8.safetensors",
+      "fp8": "diffusion_models/z-anime-distill-8step-fp8.safetensors",
+      "aio_bf16": "diffusion_models/z-anime-distill-8step-fp8.safetensors",
+      "aio_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-8step-fp8.safetensors",
+      "fp8": "diffusion_models/z-anime-distill-8step-fp8.safetensors",
+      "aio_bf16": "diffusion_models/z-anime-distill-8step-fp8.safetensors",
+      "aio_fp8": "diffusion_models/z-anime-distill-8step-fp8.safetensors",
+      "recommended_settings": {
+        "steps": 4,
+        "cfg": 1.0,
+        "sampler": "euler_ancestral",
+        "scheduler": "beta"
+      }
+    },
+    "gguf": {
+      "q8_0": {
+        "file": "gguf/z-anime-base-q4_k_s.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": "(root)",
+    "path": "diffusers/",
     "pipeline_class": "ZImagePipeline",
-    "usage": "ZImagePipeline.from_pretrained('mcuo/Anime-Z', subfolder='diffusers', torch_dtype=torch.bfloat16)",
+    "usage": "ZImagePipeline.from_pretrained('SeeSee21/Z-Anime', subfolder='diffusers', torch_dtype=torch.bfloat16)",
     "components": [
       "model_index.json",
       "scheduler/",
@@ -40,24 +81,51 @@
   "components": {
     "text_encoders": {
       "default": {
-        "bf16": "text_encoder/qwen_3_4b-bf16.safetensors",
-        "description": "Standard Z-Image text encoder, repackaged as a single safetensors"
+        "bf16": "text_encoder/qwen_3_4b-fp8.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-fp8.safetensors",
+        "fp8": "text_encoder/qwen_3_4b-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)"
+      "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/SeeSee21"
   },
   "notes": [
-    "This Space uses the distill_8step BF16 variant for fast 8-step inference.",
-    "The base pipeline is loaded via ZImagePipeline.from_pretrained() from the repo root (diffusers format).",
-    "The distilled transformer weights are then loaded separately from diffusion_models/z-anime-distill-8step-bf16.safetensors."
+    "BF16 and FP8 are the main release formats.",
+    "GGUF variants are intended for lower-memory or alternative inference setups.",
+    "AIO variants include the Text Encoder and VAE in a single file (Base, 4-Step, 8-Step).",
+    "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')."
   ]
 }