- PromptChain_ZImageRegionalCouple: $name{} per-character regional conditioning on
Z-Image's joint attention (DiT analog of AttentionCouple); base+turbo 3D-regional templates.
- Z-Image as a pickable UPSCALE engine + 'use current model' INPAINT (source path traces
CLIPLoader/lumina2); family-aware so base=dpmpp_sde/cfg4, turbo=res_multistep/cfg1.3/0.7.
- PoseStudio true-depth fix: range from posed bones (was bind-box -> figures under-exposed).
- Base model config + templates aligned to dpmpp_sde/cfg4; user-saved templates keep their
settings (lockSettings) without freezing model-swap.
- Edit-modal region-upscale: first-apply blank-layer fix (mask snapshot vs deselect race).
145 lines
3.6 KiB
JSON
145 lines
3.6 KiB
JSON
{
|
|
"display_name": "Z-Image - Base",
|
|
"model_name": "Z-Image",
|
|
"version": "Base",
|
|
"architecture": "zimage",
|
|
"family": "zimage_base",
|
|
"author": "Z-Image Team",
|
|
"description": "High-quality image generation model using Qwen text encoder. Base version — 25-50 steps, higher quality.",
|
|
"tags": [
|
|
"realistic",
|
|
"photographic",
|
|
"zimage"
|
|
],
|
|
"release_date": "2025-01-27",
|
|
"tag_sources": [],
|
|
"tag_format": "spaces",
|
|
"prompt_style": "natural",
|
|
"trigger": "4k, highly detailed",
|
|
"default_prompt_id": "zimage-photorealistic-default",
|
|
"negative": "",
|
|
"quality_position": "end",
|
|
"notes": "CFG 4.0, Steps 25-50, dpmpp_sde+beta (base is non-distilled — a stochastic sampler gives real skin texture; res_multistep is Turbo's and sterilizes it). Uses Qwen3 4B encoder, no clip skip.",
|
|
"files": [
|
|
{
|
|
"label": "Diffusion Model",
|
|
"folder": "diffusion_models",
|
|
"filename": "z_image_bf16.safetensors",
|
|
"size_bytes": 12309866400,
|
|
"source": {
|
|
"provider": "huggingface",
|
|
"repo": "Comfy-Org/z_image",
|
|
"path": "split_files/diffusion_models/z_image_bf16.safetensors"
|
|
},
|
|
"quick_hash": "2cba1b6dd3930309",
|
|
"sha256": null
|
|
},
|
|
{
|
|
"label": "Text Encoder",
|
|
"folder": "text_encoders",
|
|
"variants": [
|
|
{
|
|
"precision": "bf16",
|
|
"filename": "qwen_3_4b.safetensors",
|
|
"size_bytes": 8044982048,
|
|
"source": {
|
|
"provider": "huggingface",
|
|
"repo": "Comfy-Org/z_image",
|
|
"path": "split_files/text_encoders/qwen_3_4b.safetensors"
|
|
},
|
|
"sha256": null
|
|
},
|
|
{
|
|
"precision": "fp8",
|
|
"filename": "qwen_3_4b_fp8_mixed.safetensors",
|
|
"size_bytes": 5631994051,
|
|
"source": {
|
|
"provider": "huggingface",
|
|
"repo": "Comfy-Org/z_image",
|
|
"path": "split_files/text_encoders/qwen_3_4b_fp8_mixed.safetensors"
|
|
},
|
|
"sha256": null
|
|
},
|
|
{
|
|
"precision": "fp4",
|
|
"filename": "qwen_3_4b_fp4_mixed.safetensors",
|
|
"size_bytes": 3479416193,
|
|
"source": {
|
|
"provider": "huggingface",
|
|
"repo": "Comfy-Org/z_image",
|
|
"path": "split_files/text_encoders/qwen_3_4b_fp4_mixed.safetensors"
|
|
},
|
|
"sha256": null
|
|
}
|
|
]
|
|
},
|
|
{
|
|
"label": "VAE",
|
|
"folder": "vae",
|
|
"filename": "ae.safetensors",
|
|
"size_bytes": 335304388,
|
|
"source": {
|
|
"provider": "huggingface",
|
|
"repo": "Comfy-Org/z_image",
|
|
"path": "split_files/vae/ae.safetensors"
|
|
},
|
|
"sha256": null
|
|
}
|
|
],
|
|
"nodes": {
|
|
"KSampler": {
|
|
"steps": 50,
|
|
"cfg": 4.0,
|
|
"sampler_name": "dpmpp_sde",
|
|
"scheduler": "beta",
|
|
"steps_range": [
|
|
25,
|
|
50
|
|
],
|
|
"cfg_range": [
|
|
3.0,
|
|
5.0
|
|
],
|
|
"sampler_name_options": [
|
|
"dpmpp_sde",
|
|
"euler_ancestral",
|
|
"euler",
|
|
"dpmpp_2m",
|
|
"res_multistep"
|
|
],
|
|
"scheduler_options": [
|
|
"beta",
|
|
"simple",
|
|
"sgm_uniform",
|
|
"normal"
|
|
]
|
|
},
|
|
"EmptySD3LatentImage": {
|
|
"width": 768,
|
|
"height": 1152,
|
|
"_resolutions": [
|
|
{
|
|
"width": 768,
|
|
"height": 1152
|
|
},
|
|
{
|
|
"width": 832,
|
|
"height": 1216
|
|
},
|
|
{
|
|
"width": 1152,
|
|
"height": 768
|
|
},
|
|
{
|
|
"width": 1216,
|
|
"height": 832
|
|
},
|
|
{
|
|
"width": 1024,
|
|
"height": 1024
|
|
}
|
|
]
|
|
}
|
|
}
|
|
}
|