Models — mods registry

Diffusion Model 7 variants

Qwen Image

by Qwen / city96

Qwen Image — 20B parameter text-to-image generation model by Qwen (Alibaba).

★★★★½ 4.8 12K —

#qwen#text-to-image#20b#gguf

mods install qwen-image

Requires: qwen-image-vae qwen-image-clip

Diffusion Model 8 variants

FLUX.2 Dev

by black-forest-labs / Comfy-Org / unsloth

FLUX.2 Dev — next-generation image model from Black Forest Labs.

★★★★½ 4.9 — —

#flux2#text-to-image#high-quality#photorealistic

mods install flux2-dev

Requires: flux2-vae flux2-mistral-text-encoder

Diffusion Model 9 variants

FLUX.2 Klein 4B

by black-forest-labs / Comfy-Org / unsloth

FLUX.2 Klein 4B — the fastest model in the FLUX family.

★★★★½ 4.8 — —

#flux2#klein#text-to-image#image-editing

mods install flux2-klein-4b

Requires: flux2-vae flux2-qwen3-4b-text-encoder

Diffusion Model 15 variants

FLUX.2 Klein 9B

by black-forest-labs / unsloth

FLUX.2 Klein 9B — mid-size model in the FLUX.2 Klein family.

★★★★½ 4.7 — —

#flux#text-to-image#image-editing#gguf

mods install flux2-klein-9b

Requires: flux2-vae flux2-qwen3-8b-text-encoder

Checkpoint 20 variants

Qwen Image Edit

by Comfy-Org / QuantStack

AI-powered image editing model based on Qwen 2.5 VL architecture.

★★★★½ 4.8 275K —

#qwen#image-editing#outpainting#scene-generation

mods install qwen-image-edit

Requires: qwen-image-vae qwen-image-clip

LoRA 10 variants

Qwen Image Edit Lightning LoRA

by lightx2v

Distillation LoRA for Qwen Image Edit that reduces inference from ~50 steps to 4-8 steps.

★★★★½ 4.7 773K —

#qwen#lora#lightning#distillation

mods install qwen-image-edit-lightning

Text Encoder

FLUX.2 Mistral 3 Small Text Encoder

by Comfy-Org

Mistral 3 Small text encoder (bf16) for FLUX.2 Dev.

★★★★★ 5 — 24.0 GB

#flux2#text-encoder#mistral#dev

mods install flux2-mistral-text-encoder

Text Encoder

FLUX.2 Qwen 3 4B Text Encoder

by Comfy-Org

Qwen 3 4B text encoder for FLUX.2 Klein 4B models.

★★★★★ 5 — 9.8 GB

#flux2#text-encoder#qwen#klein-4b

mods install flux2-qwen3-4b-text-encoder

Text Encoder

FLUX.2 Qwen 3 8B Text Encoder

by Comfy-Org

Qwen 3 8B text encoder (fp8 mixed) for FLUX.2 Klein 9B models.

★★★★★ 5 — 8.5 GB

#flux2#text-encoder#qwen#klein-9b

mods install flux2-qwen3-8b-text-encoder

VAE

FLUX.2 VAE

by black-forest-labs / Comfy-Org

VAE for all FLUX.2 models (Dev, Klein 4B, Klein 9B).

★★★★★ 5 — 335 MB

#flux2#vae#bfl

mods install flux2-vae

Text Encoder 2 variants

Qwen 2.5 VL 7B Text Encoder

by Comfy-Org

Qwen 2.5 Vision-Language 7B text/vision encoder for Qwen Image Edit.

★★★★½ 4.8 275K —

#qwen#text-encoder#vision-language#clip

mods install qwen-image-clip

VAE

Qwen Image VAE

by Comfy-Org

VAE for the Qwen Image Edit pipeline. Single file, no variants.

★★★★★ 5 275K 266 MB

#qwen#vae#comfyui

mods install qwen-image-vae

Diffusion Model 7 variants

FLUX.1 Kontext Dev

by black-forest-labs / unsloth

FLUX.1 Kontext Dev — 12B parameter image editing model by Black Forest Labs.

★★★★½ 4.9 5.3M —

#flux#kontext#image-editing#image-to-image

mods install flux-kontext-dev

Requires: clip-l t5-xxl-fp8 flux-vae

Diffusion Model 8 variants

LTX-2 19B

by Lightricks

LTX-2 19B parameter video generation model. Supports text-to-video,

★★★★½ 4.8 450K —

#ltx#ltx-2#video#text-to-video

mods install ltx-2-19b

Requires: ltx-2-text-encoder ltx-2-vae

Text Encoder

LTX-2 Embeddings Connector

by Lightricks

Embeddings connector / text encoder for LTX-2 19B video model.

★★★★½ 4.7 300K 2.9 GB

#ltx#ltx-2#text-encoder#video

mods install ltx-2-text-encoder

VAE

LTX-2 Video VAE

by Lightricks

Video VAE for LTX-2 19B model. Handles video encoding/decoding.

★★★★½ 4.7 300K 2.4 GB

#ltx#ltx-2#vae#video

mods install ltx-2-vae

Diffusion Model 4 variants

LTX-Video 13B

by Lightricks

LTX-Video 13B parameter model (v0.9.8). Higher quality than the 2B version.

★★★★½ 4.7 520K —

#ltx#ltx-video#video#text-to-video

mods install ltx-video-13b

Requires: t5-xxl-fp16

Text Encoder 2 variants

UMT5-XXL Text Encoder

by Comfy-Org

UMT5-XXL multilingual text encoder for Wan 2.1/2.2 video generation models.

★★★★½ 4.8 1.5M —

#umt5#text-encoder#wan#multilingual

mods install umt5-xxl

Diffusion Model 6 variants

Wan 2.1 VACE 14B

by Wan-AI / QuantStack

Wan 2.1 VACE 14B — Video All-in-one Control & Edit model.

★★★★½ 4.8 9.6M —

#wan#wan2.1#vace#video

mods install wan21-vace-14b

Requires: umt5-xxl wan21-vae

VAE

Wan 2.1 VAE

by Wan-AI

VAE for Wan 2.1/2.2 14B video generation models. Compact 254 MB VAE

★★★★½ 4.8 1.2M 254 MB

#wan#wan2.1#wan2.2#vae

mods install wan21-vae

Diffusion Model 2 variants

Wan 2.2 I2V 14B

by Wan-AI

Wan 2.2 14B image-to-video model (high-noise expert). Converts static images

★★★★½ 4.9 720K —

#wan#wan2.2#video#image-to-video

mods install wan22-i2v-high-noise-14b

Requires: wan22-i2v-low-noise-14b umt5-xxl wan21-vae

Diffusion Model 2 variants

Wan 2.2 I2V Low Noise Expert 14B

by Wan-AI

Low-noise expert for Wan 2.2 14B image-to-video MoE architecture.

★★★★½ 4.9 600K —

#wan#wan2.2#video#image-to-video

mods install wan22-i2v-low-noise-14b

Diffusion Model 2 variants

Wan 2.2 T2V 14B

by Wan-AI

Wan 2.2 14B text-to-video model (high-noise expert). Uses MoE architecture

★★★★½ 4.9 780K —

#wan#wan2.2#video#text-to-video

mods install wan22-t2v-high-noise-14b

Requires: wan22-t2v-low-noise-14b umt5-xxl wan21-vae

Diffusion Model 2 variants

Wan 2.2 T2V Low Noise Expert 14B

by Wan-AI

Low-noise expert for Wan 2.2 14B text-to-video MoE architecture.

★★★★½ 4.9 680K —

#wan#wan2.2#video#text-to-video

mods install wan22-t2v-low-noise-14b

Diffusion Model 5 variants

Wan 2.2 TI2V 5B

by Wan-AI

Wan 2.2 hybrid text-to-video and image-to-video 5B model. Fits on 8GB VRAM

★★★★½ 4.8 920K —

#wan#wan2.2#video#text-to-video

mods install wan22-ti2v-5b

Requires: umt5-xxl wan22-vae

VAE

Wan 2.2 VAE

by Wan-AI

VAE for Wan 2.2 video generation models. New high-compression VAE

★★★★½ 4.8 850K 1.4 GB

#wan#wan2.2#vae#video

mods install wan22-vae

Checkpoint

LTX-Video 2B

by Lightricks

Efficient 2B parameter video generation model by Lightricks. Fast inference,

★★★★½ 4.6 1.4M 6.3 GB

#ltx#ltx-video#video#text-to-video

mods install ltx-video-2b

Requires: t5-xxl-fp16

Checkpoint 2 variants

FLUX.1 Dev

by black-forest-labs

High-quality text-to-image model from Black Forest Labs.

★★★★½ 4.9 2.9M —

#flux#text-to-image#high-quality#bfl

mods install flux-dev

Requires: flux-vae t5-xxl-fp16 clip-l

Checkpoint 2 variants

FLUX.1 Schnell

by black-forest-labs

Fast text-to-image model from Black Forest Labs.

★★★★½ 4.7 1.9M —

#flux#text-to-image#fast#distilled

mods install flux-schnell

Requires: flux-vae t5-xxl-fp16 clip-l

Diffusion Model

FLUX.1 Canny Dev

by black-forest-labs

FLUX.1 Canny Dev — canny edge conditioned generation model on FLUX architecture.

★★★★½ 4.7 480K 23.8 GB

#flux#controlnet#canny#edge-detection

mods install flux-canny-dev

Requires: clip-l t5-xxl-fp8 flux-vae

ControlNet

FLUX.1 Depth ControlNet

by InstantX

Depth-conditioned ControlNet for FLUX.1 Dev. Allows generating images

★★★★½ 4.5 380K 6.6 GB

#flux#controlnet#depth#instantx

mods install flux-depth-controlnet

Diffusion Model

FLUX.1 Fill Dev

by black-forest-labs

FLUX.1 Fill Dev — inpainting and outpainting model based on FLUX architecture.

★★★★½ 4.8 650K 23.8 GB

#flux#inpainting#outpainting#fill

mods install flux-fill-dev

Requires: clip-l t5-xxl-fp8 flux-vae

IP-Adapter

FLUX.1 Redux Dev

by black-forest-labs

FLUX.1 Redux Dev — image variation adapter for FLUX. Takes image input and

★★★★½ 4.6 320K 129 MB

#flux#ipadapter#image-variation#redux

mods install flux-redux-dev

Text Encoder

CLIP-L Text Encoder

by comfyanonymous

CLIP-L (Large) text encoder. Used as secondary text encoder by FLUX models

★★★★★ 5 2.6M 246 MB

#clip#text-encoder#flux#lightweight

mods install clip-l

VAE

FLUX VAE (ae.safetensors)

by black-forest-labs

The VAE used by all FLUX models. Required for FLUX.1 Dev and Schnell.

★★★★★ 5 3.2M 335 MB

#flux#vae#bfl

mods install flux-vae

Text Encoder

T5-XXL Text Encoder (fp16)

by comfyanonymous

T5-XXL text encoder in fp16 precision. Required by FLUX models

★★★★★ 5 2.4M —

#t5#text-encoder#flux#fp16

mods install t5-xxl-fp16

Text Encoder

T5-XXL Text Encoder (fp8)

by comfyanonymous

T5-XXL text encoder quantized to fp8 precision. Uses half the VRAM

★★★★½ 4.8 1.8M —

#t5#text-encoder#flux#fp8

mods install t5-xxl-fp8

IP-Adapter

IP-Adapter FaceID (SDXL)

by h94

IP-Adapter with InsightFace face ID embedding for SDXL.

★★★★½ 4.5 800K 1.1 GB

#sdxl#ipadapter#faceid#face-consistent

mods install ip-adapter-faceid-sdxl

IP-Adapter

IP-Adapter SDXL (ViT-H)

by h94

IP-Adapter for SDXL using ViT-H image encoder. Enables image-prompted

★★★★½ 4.7 1.2M 732 MB

#sdxl#ipadapter#image-prompt#style-transfer

mods install ip-adapter-sdxl

LoRA

LCM LoRA (SD 1.5)

by latent-consistency

Latent Consistency Model LoRA for SD 1.5. Enables fast 2-8 step inference

★★★★½ 4.5 1.8M 134 MB

#sd15#lora#lcm#fast-inference

mods install lcm-lora-sd15

LoRA

LCM LoRA (SDXL)

by latent-consistency

Latent Consistency Model LoRA for SDXL. Enables fast 2-8 step inference

★★★★½ 4.6 2.2M 394 MB

#sdxl#lora#lcm#fast-inference

mods install lcm-lora-sdxl

Checkpoint

Stable Diffusion XL Base 1.0

by stabilityai

Stable Diffusion XL base model. High resolution text-to-image generation

★★★★½ 4.6 5.4M —

#sdxl#text-to-image#stable-diffusion#high-resolution

mods install sdxl-base-1.0

Requires: sdxl-vae-fp16-fix

ControlNet 2 variants

ControlNet Canny (SDXL)

by diffusers

ControlNet trained on SDXL for Canny edge detection conditioning.

★★★★½ 4.6 1.5M —

#sdxl#controlnet#canny#edge-detection

mods install sdxl-controlnet-canny

Checkpoint

SDXL Refiner 1.0

by stabilityai

SDXL 1.0 Refiner — small-detail expert model. Used as a second pass

★★★★½ 4.5 1.8M 6.5 GB

#sdxl#refiner#detail#stable-diffusion

mods install sdxl-refiner-1.0

Requires: sdxl-vae-fp16-fix

Checkpoint 2 variants

SDXL Turbo

by stabilityai

SDXL Turbo — distilled from SDXL 1.0 using Adversarial Diffusion Distillation.

★★★★½ 4.7 2.8M —

#sdxl#turbo#fast#text-to-image

mods install sdxl-turbo

Requires: sdxl-vae-fp16-fix

Segmentation

BiRefNet DIS (Dichotomous Image Segmentation)

by ViperYX

High-quality background removal / image segmentation model.

★★★★½ 4.8 65K 889 MB

#segmentation#background-removal#birefnet#matting

mods install birefnet-dis

Upscaler

4x UltraSharp Upscaler

by Kim2091

High-quality 4x upscaler. Excellent for photo-realistic upscaling.

★★★★½ 4.9 3.2M 67 MB

#upscaler#4x#photo-realistic#esrgan

mods install 4x-ultrasharp

Upscaler

RealESRGAN x4plus

by xinntao

General-purpose 4x upscaler from the Real-ESRGAN project.

★★★★½ 4.7 4.5M 67 MB

#upscaler#4x#realesrgan#general-purpose

mods install realesrgan-x4plus

Checkpoint

Stable Diffusion 1.5

by runwayml

The original Stable Diffusion 1.5. Lightweight, fast, huge ecosystem

★★★★☆ 4.3 12.0M —

#sd15#text-to-image#stable-diffusion#lightweight

mods install sd-1.5

Requires: sd-vae-ft-mse

Checkpoint

Stable Diffusion 2.1

by stabilityai

Stable Diffusion 2.1, fine-tuned from SD 2.0 with improved aesthetics.

★★★★½ 4.4 3.2M 5.2 GB

#sd21#text-to-image#stable-diffusion#768

mods install sd-2.1

VAE

SD VAE ft-MSE

by stabilityai

Fine-tuned VAE for Stable Diffusion 1.5. Produces sharper, more

★★★★½ 4.7 6.8M 335 MB

#sd15#vae#fine-tuned#mse

mods install sd-vae-ft-mse

ControlNet

ControlNet v1.1 Canny (SD 1.5)

by lllyasviel

ControlNet v1.1 for SD 1.5 — Canny edge detection conditioned generation.

★★★★½ 4.8 4.2M 1.6 GB

#sd15#controlnet#canny#edge-detection

mods install sd15-controlnet-canny

ControlNet

ControlNet v1.1 Depth (SD 1.5)

by lllyasviel

ControlNet v1.1 for SD 1.5 — Depth map conditioned generation.

★★★★½ 4.7 3.5M 1.6 GB

#sd15#controlnet#depth#midas

mods install sd15-controlnet-depth

ControlNet

ControlNet v1.1 OpenPose (SD 1.5)

by lllyasviel

ControlNet v1.1 for SD 1.5 — OpenPose body pose conditioned generation.

★★★★½ 4.7 3.1M 1.6 GB

#sd15#controlnet#openpose#pose

mods install sd15-controlnet-openpose

VAE

SDXL VAE (fp16 NaN fix)

by madebyollin

Fixed SDXL VAE that works correctly in fp16 precision.

★★★★½ 4.9 4.1M 335 MB

#sdxl#vae#fp16-fix

mods install sdxl-vae-fp16-fix

Upscaler

BSRGANx2 Upscaler

by cszn

2x upscaler based on BSRGAN (Blind Super-Resolution GAN).

★★★★½ 4.6 450K 72 MB

#upscaler#2x#bsrgan#denoising

mods install bsrganx2

Text Encoder 3 variants

Z-Image Qwen 3 4B Text Encoder

by Comfy-Org

Qwen 3 4B text encoder used by Z-Image-Turbo.

★★★★½ 4.8 — —

#z-image#text-encoder#qwen3#comfyui

mods install z-image-text-encoder

Diffusion Model 10 variants

Z-Image-Turbo

by Tongyi-MAI / Comfy-Org / jayn7

Distilled 6B parameter text-to-image model from Alibaba Tongyi Lab.

★★★★½ 4.9 — —

#z-image#text-to-image#turbo#fast-inference

mods install z-image-turbo

Requires: z-image-text-encoder z-image-vae

ControlNet

Z-Image-Turbo Fun ControlNet Union

by alibaba-pai

ControlNet union model for Z-Image-Turbo.

★★★★½ 4.6 — 3.3 GB

#z-image#controlnet#canny#depth

mods install z-image-turbo-controlnet-union

Requires: z-image-turbo

LoRA

Z-Image-Turbo Distill Patch LoRA

by Comfy-Org

Distillation patch LoRA for Z-Image-Turbo.

★★★★½ 4.7 — 167 MB

#z-image#lora#distillation#comfyui

mods install z-image-turbo-distill-lora

VAE

Z-Image VAE

by Comfy-Org

VAE for Z-Image-Turbo. Decodes latents to images.

★★★★★ 5 — 351 MB

#z-image#vae#comfyui

mods install z-image-vae