Stable Video Diffusion ComfyUI Install

An easier way to generate videos using stable video diffusion models.

Stable Video Diffusion ComfyUI install:

Requirements:

you can use any safetensor file, I used the base SDXL model.

Instructions:

- Download the ComfyUI portable standalone build for Windows

- Install ComfyUI-Manager (optional)

- Install VHS - Video Helper Suite (optional)

- Download either of the .safetensors from to the "ComfyUI-checkpoints" -folder

- Load JSON file

- Best settings to use are:
- 1024w x 576h
- num frames for svd: 14
- num frames for svd_xt: 25

[FightClubGarmz]

This is the first tutorial that everything worked first time, very happy with the length and pace of the tutorial. You have now got your self a sub, as well as a very grateful advocate of your video. Thank you, I will go and explore how to hone this into making the animations I have to create. Thank you

[MrStetsonator]

Excellent explainer, thanks for this! Hadn't used ComfyUI prior to this and got video working in a half hour

[KardinalMoses]

I appreciated how you explained what each step does and what the differences are. Everything was very understandable and easy to implement. Thank you very much :)

[korbinianvogt]

Fantastic tutorial! Actually like Pika and Runway more so far, as it animates the person better or more often, but nice to have it set up for future updates of SDV!

[jacob-ydhy]

Please keep updating anything SVD related on this channel. Valuable info!

[ESGamingCentral]

This method worked, I have been using the new models without issue. I'm still playing with the settings.

[MrPlasmo]

super helpful tutorials ! One minor suggestion would be to record in 4k as some of the text is a little blurry on 1080p with my old eyes :) thanks again

[AlexisLionel]

Great tutorial, man! Thank you very much!

[dustboii]

Thanks again. This was surprisingly simple. Im curious can you just generate at a higher resolution or is there a upscaling node we can use? If so can you do a short video on it.

[realaipirate]

How would you recommend the best way to getting more control over the movement? Outside of scaling the motion bucket, I'd like to add camera movements or prompt the movements.

[tailongjin-yxki]

is it suitable for mac? i tried lot but can't make it

[realaipirate]

Some of these nodes aren't loading up, would you mind dropping a link to the "SVD_img2vid_conditioning", "VideoLinearCFGGuidance", and "image only checkpoint loader (img2vid model)" I searched for them on the Comfyui Manager and they don't show up there. -- thank you for this tutorial!

[Tarantella.Serpentine]

Great tutorial everything works! which is saying alot lol... no Cuda errors or path errors or pip errors I could go on... :) Can you do a Video on merging datasets for a frankenstein LLM :)

[nicktumi]

This is awesome! Thank you! I really like this format for development with comfy. Are there other AI's that can be utilized in this IDE? I'm looking at trying StableSR for upscaling old photos, ideally, I'd like a MAGNIGIC AI quality, but from my research that is a subscription model. Also would really like to install Spleeter in ComfyUI.

[dustboii]

For some reason the comfy UI wont extract from the zip for me. If I drag the folder to user folder it just sits at zero. If I try to hit extract all on the zip it gives errors. EDIT... Oh I guess you literally need 7zip. Weird.

[Sususuck]

First of all, I want to thank you for providing this method. However, I've encountered a small issue... When I use MP4 as the output method, the first half of the video always appears completely black?

[deimoslyric]

Thank you so much! However I am having the issue when trying to add VHS as a node. Neither VHS nor Videocombine come up as an option, and I do not know how to get it to come up.

[racostitum]

Great video. Thank you!
I have tried to install VHS_VideoCombine but some sort of error comes up and says IMPORT FAILED what to do?

[amkkart]

Do I have install comfyui again or can I load the model into my existing comfyui?

[SwenstrOOp]

Hi, thanks for tutorial, do you know how to save each frame of video as png in this workflow?