DreamX Creator Comfy

ComfyUI-ready DreamX-Creator generator, Audio VAE, Wan dependencies, and causal 2× refiner weights.

Install for ComfyUI

Install the hf command if needed:

python -m pip install -U huggingface_hub

Download this repository directly into ComfyUI's shared model directory:

hf download t8star/DreamX-Creator-Comfy --local-dir ComfyUI/models/dreamx_creator

The resulting model root must be:

ComfyUI/
└── models/
    └── dreamx_creator/          # select model_root=auto in the loader
        ├── creator/
        │   ├── cross_attn_weights.safetensors
        │   ├── merged_lora_info.json
        │   ├── audio_model/
        │   │   ├── config.json
        │   │   └── diffusion_pytorch_model.safetensors
        │   └── video_model/
        │       ├── config.json
        │       ├── diffusion_pytorch_model-00001-of-00002.safetensors
        │       ├── diffusion_pytorch_model-00002-of-00002.safetensors
        │       └── diffusion_pytorch_model.safetensors.index.json
        ├── audio_vae/
        │   ├── config.json
        │   └── diffusion_pytorch_model.safetensors
        ├── refiner/
        │   ├── sr_dit_5b.pt
        │   ├── latent_upsampler_flash.pt
        │   ├── latent_upsampler_2d_causal.pt
        │   └── lightvae_nu_scheme3.pt
        └── wan2.2_ti2v_5b/
            ├── Wan2.2_VAE.pth
            ├── models_t5_umt5-xxl-enc-bf16.pth
            └── google/
                └── umt5-xxl/
                    ├── special_tokens_map.json
                    ├── spiece.model
                    ├── tokenizer_config.json
                    └── tokenizer.json

model_root=auto searches the custom-node repository's checkpoints/ first, then ComfyUI/models/dreamx_creator/. Therefore this node-local layout also works:

hf download t8star/DreamX-Creator-Comfy --local-dir ComfyUI/custom_nodes/Comfyui-DreamX-Creator-T8/checkpoints

The complete bundle contains 20 model/config/tokenizer files and is approximately 54.25 GB (50.53 GiB). Validate the download before loading the models:

python ComfyUI/custom_nodes/Comfyui-DreamX-Creator-T8/scripts/verify_models.py ComfyUI/models/dreamx_creator

Components

Directory	Used by	Contents
`creator/`	Creator workflow	7B joint video/audio DiTs and cross-modal attention
`audio_vae/`	Creator workflow	CreatorDACVAE audio decoder
`refiner/`	Refiner workflow	SR-DiT 5B and latent upsamplers
`wan2.2_ti2v_5b/`	Both workflows	Wan VAE, UMT5-XXL encoder, and tokenizer

The Creator and Refiner are intended to run as separate ComfyUI workflow phases so both large models do not remain resident on the GPU at the same time.

T8star social links

Resource	Link
Bilibili	T8star on Bilibili
YouTube	@T8star-Aix
Seedance API	API signup
Online AI Apps	RunningHub profile
ComfyUI Package	Quark download
Hugging Face profile	huggingface.co/t8star

Attribution and license

These files are organized for the native ComfyUI nodes from the weights released by the DreamX Team at GD-ML/DreamX-Creator. The project is licensed under Apache License 2.0. See the LICENSE file in this repository. Please retain the original attribution when redistributing the weights.

@misc{zhu2026dreamxcreatordemocratizingnativeaudiovideo,
  title={DreamX-Creator: Democratizing Native Audio-Video Generation at 2K Resolution},
  author={Jiashu Zhu and Yanhao Zheng and Ruitian Tian and Rujing Dang and Shen Zhang and Bingze Song and Jiachen Lei and Ruimin Lin and Jiahong Wu and Xiangxiang Chu},
  year={2026},
  eprint={2608.31106},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2608.31106}
}

t8star/DreamX-Creator-Comfy