greed (Z-Image base finetune) — ComfyUI format

Diffusion model (DiT, denoiser only) finetuned from Z-Image (Alibaba Tongyi, S3-DiT), converted from a Draw Things export to ComfyUI-compatible safetensors.

• gr33d_main_model_full_f16_comfyui.safetensors — FP16, ~11GB

Usage (ComfyUI)

Place in ComfyUI/models/diffusion_models/, load with Load Diffusion Model (UNETLoader). This is the diffusion model only; you also need:

• Text encoder: Qwen3-4B (qwen_3_4b.safetensors) in models/text_encoders/
• VAE: Flux VAE (ae.safetensors, 16-channel) in models/vae/

Use a standard Z-Image (AuraFlow-style) workflow.

Conversion notes

Key-remapped from Draw Things layout to Z-Image/NextDiT naming (fused qkv, concatenated adaLN modulation, flattened norm dims). Verified: all 453 tensor keys and shapes match the official Comfy-Org Z-Image diffusion model exactly.