ImageGen

Models

Models are not included. This list says which file belongs in which folder under ${IMAGEGEN_DATA}/models/ and where it comes from. You only need the models of the workflows you want to use. The model card of the source is always authoritative for the license; the information here is a help, not legal advice.

Only download .safetensors: .ckpt/.pt can execute code when loaded.

Workflow names as shown in the English interface.

Images

Workflows	File	Folder	Source	License
Realistic, Painted - D&D (illustration, location, token, token from image), Edit, Character	`qwen_image_2.1_turbo_bf16.safetensors` (14.2 GB)	`diffusion_models/`	Comfy-Org/Qwen-Image-2.1	Qwen Research License – non-commercial
	`qwen3vl_8b_bf16.safetensors` (17.5 GB)	`text_encoders/`	same	same
	`qwen_image_2.1_vae_bf16.safetensors`	`vae/`	same	same
Anime - Anima	`anima-turbo-v1.1.safetensors`	`diffusion_models/`	circlestone-labs/Anima (`split_files/`)	see model card
D&D battle map, location map	`anima-aesthetic-v1.1.safetensors`	`diffusion_models/`	same	see model card
Anima, battle map, location map	`qwen_3_06b_base.safetensors`	`text_encoders/`	same	see model card
	`qwen_image_vae.safetensors`	`vae/`	same	see model card
D&D battle map, location map	`mapcraft_anima_v1.safetensors` (LoRA)	`loras/`	Civitai, model 799901 (version for Anima)	terms on the model page
Test - Flux 3 Klein	`flux3-image-klein-bf16.safetensors` (11.1 GB)	`diffusion_models/`	black-forest-labs/flux-3-image-klein-comfy (accept the license on the page, then download with a token)	FLUX Community License
	`qwen3vl_4b_bf16.safetensors`	`text_encoders/`	same	same
	`flux2-vae.safetensors`	`vae/`	same	same

Videos (Wan 2.2)

About 69 GB in total. Source: Comfy-Org/Wan_2.2_ComfyUI_Repackaged (split_files/), text encoder and VAE from Comfy-Org/Wan_2.1_ComfyUI_repackaged. License: Apache-2.0; for the Lightx2v LoRAs see the model card of lightx2v.

Workflows	File	Folder
Video	`wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors`, `wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors`	`diffusion_models/`
Video	`wan2.2_t2v_lightx2v_4steps_lora_v1.1_high_noise.safetensors`, `..._low_noise.safetensors`	`loras/`
Image to video (both)	`wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors`, `wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors`	`diffusion_models/`
Image to video (fast)	`wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors`, `..._low_noise.safetensors`	`loras/`
all videos	`umt5_xxl_fp8_e4m3fn_scaled.safetensors`	`text_encoders/`
all videos	`wan_2.1_vae.safetensors`	`vae/`

Your own extensions

LoRAs that an extension in extras/ adds (see workflows/extras/README.en.md) also go into loras/. The terms of their source apply to them.

Downloading

With the browser or e.g.:

cd "${IMAGEGEN_DATA:-$HOME/imagegen}/models"
curl -L -o diffusion_models/qwen_image_2.1_turbo_bf16.safetensors \
  https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/diffusion_models/qwen_image_2.1_turbo_bf16.safetensors

For gated models (FLUX 3) add -H "Authorization: Bearer ". Then press “R” (refresh) in ComfyUI or reload the page.

View on GitHub → · Project page →

Models

#Images

#Videos (Wan 2.2)

#Your own extensions

#Downloading

Images

Videos (Wan 2.2)

Your own extensions

Downloading