Ooh who's this cutie?

n_n said:
Ooh who's this cutie?

Geck the moth-gecko O.O

jelloponies said:
Geck the moth-gecko O.O

Makes sense XD

He's just adorable 🥰

Husband

Yum-my! A goat spit-roast would sure hit the spot :P

Man, what a feast for the eyes.

For herbivores, those bois sure crave being stuffed with meat.

snowwolf-roukan said:
Well yeah. I was meaning like site? Program? I'm tryin to learn this stuff ^_^; Sorry I should of been more clear.

local gen minimax h3, using fl2v as a ref model.
There are quite a lot of nuances, but you can look for answers in the poster older videos that he posted to my similar question.

Ooh~ this small boi was hiding such a huge package~ :3

Hey there,

love the animation. Do you gen the whole 1:40 in one go (10sec takes me like 40min) or do you just make a 5sec animation (5sec takes me like 10-15min) and use the last frame as the start for the new 5sec video and stitch them all together in a video processing program?

Thanks in advance

That is a really great animation; what I liked was when Vulpix quickly moved its tail back and forth, and the kiss on the forehead.

snowwolf-roukan said:
Well yeah. I was meaning like site? Program? I'm tryin to learn this stuff ^_^; Sorry I should of been more clear.

TemtemLover pretty much covered it. I do use loras as well, and ..

psycho1712 said:
...Do you gen the whole 1:40 in one go (10sec takes me like 40min) or do you just make a 5sec animation...

My process is basically: get the references ready and write the prompts for 3-5 projects at once, 20 to 22 second clips which take about ~30 minutes on average per clip, and then get them all running on multiple seeds before I head to work/sleep. Then when I get back, look at what scenes generated well enough to advance from, and continue. If a scene gets really close but not quite, I'll often move an image from it to novel ai or stable diffusion, in paint it to be right, then use it as a reference for minimax for the next Gen (e.g. "This part of the scene should look like <Picture 6>").

The seed hunter workflow stitches scenes together automatically when using a video reference for the last scene (it uses the last moments of the video before it as part of the input) but I only allow it to stich 3 20s scenes at a time, as quality degrades after that point. I use Shotcut for the final edit/putting it all together.

malico said:
TemtemLover pretty much covered it. I do use loras as well, and ..

My process is basically: get the references ready and write the prompts for 3-5 projects at once, 20 to 22 second clips which take about ~30 minutes on average per clip, and then get them all running on multiple seeds before I head to work/sleep. Then when I get back, look at what scenes generated well enough to advance from, and continue. If a scene gets really close but not quite, I'll often move an image from it to novel ai or stable diffusion, in paint it to be right, then use it as a reference for minimax for the next Gen (e.g. "This part of the scene should look like <Picture 6>").

The seed hunter workflow stitches scenes together automatically when using a video reference for the last scene (it uses the last moments of the video before it as part of the input) but I only allow it to stich 3 20s scenes at a time, as quality degrades after that point. I use Shotcut for the final edit/putting it all together.

Thanks for the reply and the information, will look into some things you wrote. :-)

technical-grid said:
Noooo!! They glued the squirrels together 😭

It's sticky for sure, but hopefully not that sticky. (Some of the alternatives generated were even more glue-like though.)

Must have balls the size of gophers.

What model/lora do I need for this?

kerewa said:
What model/lora do I need for this?

I don't wanna be too specific, but it's just one of the mainstream online tools, where I phrased the prompt carefully by trial and error to avoid triggering the filters.

He nutted alright

Some tourist spilled a condensed milk on a tree, but fellow squirrels don't mind it at all.