3D canvas move on Stable Diffusion animation

fabiyamada(67)

Published in

Alien Art Hive

Words

381

Reading

2 min

Listen

Play

For this tutorial you will need Stable Diffusin and Deforum addon, in case you don't have them, check this post first: Tutorial: USING STABLE DIFFUSION TO CREATE A PROMPT TO PROMPT ANIMATION.

I will use the following specs on Run tab:

I changed the width and height to better match a portrait image, don't forget to use the sliders. Decreased steps to 20 so this takes a bit shorter; and checked Restore faces to have better results on the person's face we will generate... guess who!

These are the prompts I will go with:

{
    "0": "henry cavill as superman, epic pose, cyberpunk, futuristic, city, epic scenario, night ",
    "30": "henry cavill as medieval knight with futuristic armor and electronic gadgets",
    "60": "henry cavill as astronaut with cowboy hat, western, cyberpunk, night"
}

Write whatever you want on the Prompts tab:

Now go to Keyframes tab and check 3D:

Scroll down a bit and you will find you can move the canvas in the Motion tab, for this test I will only add movements on Translation Z and Rotation 3D Y.

Also for this tests I am creating a file where I am saving my settings in case I want to use them again, on the right you only need to type a name and then click on Save settings. Make sure to use a different name or you will overwrite and existing file. For loading only type the name of the file and click on Load settings.

Seems like we're ready to go, so let's try!

I really love this superman suit tho!

And this is what we got... so I have to complain this handsome guy doesn't look like Henry Cavill, but he's not baad... now let's see what happens if we make some changes on the motion specs. This time we try to be a bit more subtle. And also hope we get a Henry Cavill...

And this is our amazing new output:

I don't know who this model thinks Henry Cavill is but I wouldn't mind seeing this guy as Superman...

Speaking of model, I am using revAnimated_v122; and I want to give a try using RealisticVision_V51, this one claims to be well trained on celebrities so let's see what we get...

This one looks a bit more like Henry Cavill if you ask me. So now just for the funs I will use Arcane diffusion model and modify a little the canvas move...