ComfyUI: Stable Video Diffusion (Workflow Tutorial)

This is a comprehensive workflow tutorial on using Stable Video Diffusion in Comfy UI. Stable Video Weighted Models have officially been released by Stabality AI and support up to 25 frames per second of video generation. While it might seem that the motion generation is random, it is not. In the tutorial, I showcase 6 unique workflow examples explaining each and every setting regarding the new video nodes and how to gain control over the motion animation.

------------------------

Relevant Links:

------------------------

TimeStamps:

0:00 Intro.
1:02 Requirements.
4:29 Candle Workflow.
14:58 Hand Waving Workflow.
18:12 Portrait Workflow.
29:46 Cycling Workflow.
31:34 Motobike Workflow.
33:39 Boat Workflow.
43:00 Text2Image2Video Workflow.

[controlaltai]

For Stable Video Diffusion XT 1.1, these settings work great:

Motion Bucket - 127
Min CFG 2
Target CFG 2.5
Augmentation (0 - 0.10) - mostly 0
Steps 40
Sampler Euler Ancestral
Scheduler SGM_uniform

[Adrian_Marmy]

This must be one of the best tutorials I've ever seen. Not too slow, straight to the point and still very clear. 👌

[ppaayyaamm]

Awesome / Best explanation Of SVD is Here .. Good lock

[tomaszwota]

A tip: You can get rid of manually setting the resolution in SVD_img2vid_Conditioning by passing the cropped image through "Generation Resolution From Image" which takes an image and spits out width and height -- connect those to the width and height converted into inputs in SVD img2vid Conditioning and voila, you can push different resolution images into it and it will automatically work fine.

[gabrielmorod]

This is beautiful. Thank you so much for explaining everything in detail!

[jagsdesign]

super cool explanation and methods and very methodical approach to a creative journey

[jeffscott]

Great video! Detailed explanation of the nodes, adjustments and workflows helped a lot.

[dariayudina]

this is the most useful video about comfy ui animation that i've already seen. thank you!

[tcmbravo]

Excellent tutorial so far in internet. Keep going

[manfactory]

Thank you so much for providing this VERY helpful tutorial.

[ryanstark]

Great job! So clear description, it's very helpful. Thanks for your work.

[joshualloyd]

Love this. You have made my last week or so (very new to this) so amazing! I can not thank you enough! Please don't stop! We need you...

[PirateWasHere]

This is GOLD !, I have to say, U re special ! Thanks for step by step 🍻

[oracleofwater]

Memberships, neat. That's a good way to make sure you never see any of my money.

[doanyai]

Thank you very much！Your teaching quality is very high！ If we use the LCM model and VAE, will it be faster?

[manolomaru]

Suuuuper and tremendous video!
...Simply wow
✨✨

[danilsi]

🐌The second one! As always a fascinating guide with detailed explanations. Let's practice🧑‍🔬. Thank you very much for your video and the channel in general.

[perspectalog]

Amazing as always Have you tried to build a SVD inpaint workflow?

[revidcuteo]

What you convey is really helpful, I'll subscribe for that. I have a small question for you: how can we increase the duration of an image after processing? For example, if I have an image that lasts 2 seconds after processing, how can I extend it to 4 seconds or more?

[saffyk]

Hey amazing video very educational!! Do you have plans to make one for animatediff sparse ctrl? I have tried both and sparse ctrl is not bad at all, it cuts down the quality a little bit but i still cant figure out how to control the motion