Update ComfyUI and all nodes to the latest version.

Version 5c

SageAttention and TeaCache are included.

This workflow takes about 2 minutes to generate a 2-second video.

(based on RTX 3090)

You shouldn't expect high-quality video generation from the start.

Creating a high-quality video requires adjusting many parameters, which can involve a lot of trial and error.

Requirements :

• rgthree-comfy

• ComfyUI-GGUF

• ComfyUI-KJNodes

• ComfyUI-TeaCache

• ComfyUI-Impact-Pack

• ComfyUI-Image-Filters

• comfyui-custom-scripts

• ComfyUI-VideoHelperSuite

• comfyui-frame-interpolation

• ComfyUI-HunyuanVideoMultiLora

If you are using SageAttention in a Windows environment, go to the URL and configure the settings.

URL:https://old.reddit.com/r/StableDiffusion/comments/1h7hunp/how_to_run_hunyuanvideo_on_a_single_24gb_vram_card/

UPSCALE MODEL: 4x_foolhardy_Remacri install from ComfyUI Manager -> Model Manager

MODEL: https://huggingface.co/Kijai/SkyReels-V1-Hunyuan_comfy/tree/main

Select any model download and Put models/diffusion_models/

VAE: https://huggingface.co/Kijai/HunyuanVideo_comfy/blob/main/hunyuan_video_vae_bf16.safetensors

Put models/vae/

CLIP 1: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors

Put models/clip/

CLIP 2: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors

Put models/clip/

Version 3 (Q4)

If your VRAM is 16GB or less, try reducing the number of frames.

If the results are not good, you can try the following:

1. Change the control_after_generate of the Pass 1 node to increment or randomize.

2. Modify the prompt or use a different prompt

3. Adding "FPS-24, " to the prompt

4. Increase the resolution a little

5. Increase the steps

List used:

comfyui-logic

rgthree-comfy

ComfyUI-GGUF

ComfyUI-KJNodes

comfyui-custom-scripts

ComfyUI-VideoHelperSuite

comfyui-frame-interpolation

ComfyUI-HunyuanVideoMultiLora

MODEL: https://huggingface.co/Kijai/SkyReels-V1-Hunyuan_comfy/blob/main/skyreels-hunyuan-I2V-Q4_K_S.gguf

Put models/diffusion_models/

VAE: https://huggingface.co/Kijai/HunyuanVideo_comfy/blob/main/hunyuan_video_vae_bf16.safetensors

Put models/vae/

CLIP 1: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors

Put models/clip/

CLIP 2: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors

Put models/clip/

Version 2 (Q5)

ComfyUI-GGUF

comfyui-logic

ComfyUI-KJNodes

ComfyUI-VideoHelperSuite

comfyui-florence2

comfyui-custom-scripts

comfyui-frame-interpolation

MODEL: https://huggingface.co/Kijai/SkyReels-V1-Hunyuan_comfy/blob/main/skyreels-hunyuan-I2V-Q5_K_M.gguf

Put models/diffusion_models/

VAE: https://huggingface.co/Kijai/HunyuanVideo_comfy/blob/main/hunyuan_video_vae_bf16.safetensors

Put models/vae/

CLIP 1: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors

Put models/clip/

CLIP 2: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp8_scaled.safetensors

Put models/clip/

Version 1

32GB RAM / 24GB VRAM

ComfyUI-KJNodes

ComfyUI-VideoHelperSuite

comfyui-florence2

comfyui-custom-scripts

comfyui-frame-interpolation

MODEL: https://huggingface.co/Kijai/SkyReels-V1-Hunyuan_comfy/blob/main/skyreels_hunyuan_i2v_bf16.safetensors

Put models/diffusion_models/

VAE: https://huggingface.co/Kijai/HunyuanVideo_comfy/blob/main/hunyuan_video_vae_bf16.safetensors

Put models/vae/

CLIP 1: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/clip_l.safetensors

Put models/clip/

CLIP 2: https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/blob/main/split_files/text_encoders/llava_llama3_fp16.safetensors

Put models/clip/