"A.I." (Machine Learning Algorithms) To Generate Art

MrStochastic

@AudioGus said:
This stuff is evolving pretty quick and there is a ton of noise now.

@kirmesteggno said:
This video helped me to set it up: YT Link

Thanks to you both for your helpful suggestions. I'll be trying to keep up with it all. I've had some mixed results uploading my own images to Dall-E. But sometimes it has really surprised me by keeping it vague and then seeing how it interprets indistinct images.

I've been trying out Pikaso which has a different sort of tool set and seems aimed more for graphic design type of work: https://www.freepik.com/ai/pikaso-ai-drawing?ref=futuretools.io

kirmesteggno

@MrStochastic said:
I've been trying out Pikaso which has a different sort of tool set and seems aimed more for graphic design type of work: https://www.freepik.com/ai/pikaso-ai-drawing?ref=futuretools.io

Their onboarding sucks though, wanted to try it out but while it was drawing it triggered an unihidable optin pop-up. I'll never visit that site again.

MrStochastic

@kirmesteggno said:

@MrStochastic said:
I've been trying out Pikaso which has a different sort of tool set and seems aimed more for graphic design type of work: https://www.freepik.com/ai/pikaso-ai-drawing?ref=futuretools.io

Their onboarding sucks though, wanted to try it out but while it was drawing it triggered an unihidable optin pop-up. I'll never visit that site again.

true, it doesn't give you much time to try for free; not sure, maybe 10 min. a day? but if you can tolerate that, you can import or paint/draw your own 'prompts'. Then there's a button somewhere you click and it just spits out dozens of variations at once that you can see on one page and download each one.

kirmesteggno

@MrStochastic said:

@kirmesteggno said:

@MrStochastic said:
I've been trying out Pikaso which has a different sort of tool set and seems aimed more for graphic design type of work: https://www.freepik.com/ai/pikaso-ai-drawing?ref=futuretools.io

Their onboarding sucks though, wanted to try it out but while it was drawing it triggered an unihidable optin pop-up. I'll never visit that site again.

true, it doesn't give you much time to try for free; not sure, maybe 10 min. a day? but if you can tolerate that, you can import or paint/draw your own 'prompts'. Then there's a button somewhere you click and it just spits out dozens of variations at once that you can see on one page and download each one.

For me it was way less than 10min, as soon as I moved my mouse to make a stroke it popped up, meaning I couldn't really try it out at all without signing up. They do this onboarding interruption shit because it works for tech unsavvy people they can track. (the rest won't show up in their a/b test statsistics).

I'm not against signups or paying for apps, it's just the way they do it what annoys me.

I think Leonardo.ai is a much better choice. For business graphics Canva is quite nice, their paid plan also includes an AI and the're many AI plugins for it as well. Last time I've used it was based on Dall-E which also generates 4 variants to pick from. I guess it's also what freepik/Pikaso is using under the hood.

MrStochastic

@kirmesteggno said:
I think Leonardo.ai is a much better choice.

I will give that a try. It does look good.

ChrisG

If you have a fast iPad, get Draw Things from the AppStore (completely free). It does everything Stable Diffusion, locally on your device. You can download different models in-app (free obviously), or from sites like CivitAI, as well as Lora’s, controlnets etc etc is fully supported. SDV also supported, but only good to make gifs with (25 images max).

ChrisG

Perhaps should’ve used a more family friendly gif oh well

NeuM

It's going to be really interesting when Sora is finally released. Video and even movies will never be made quite the same way again.

https://openai.com/sora

ChrisG

Looking forward to a version of Sora that runs ok on an iPad, or an iPad that runs Sora ok.

SVD is a mess. You don’t give it a prompt, you just give it a picture and you can change a few parameters that no one really knows the science behind.

The above two turned out ok, this one not so much.

kirmesteggno

I wonder if the whole extrapolation thing (e.g. from quick drawings to images) also comes for video. You've likely seen movie sets like Lord of the Rings that was built in miniature, or where chicken movements are tracked for dino models in making offs.

Building low fidelity scenes with toys and/or play dough and converting them into movies could become a cool parent child activity.

NeuM

The crowning achievement of the craft of A.I. filmmaking to date ...

AudioGus

Flux is here! Open source ass kicker...

Best to start here...
https://replicate.com/black-forest-labs/flux-schnell

https://replicate.com/black-forest-labs

AudioGus

AudioGus

AudioGus

AudioGus

NeuM

A lot of exciting developments in both still images and generated video recently. Nice examples, @AudioGus.

u0421793

I posted this yesterday
https://toot.pikopublish.ing/notes/9wkve0zxbwjadsqh

NeuM

@u0421793 said:
I posted this yesterday
https://toot.pikopublish.ing/notes/9wkve0zxbwjadsqh

I love that classic rock band, "Clasc Roltiock".

u0421793

@NeuM said:

@u0421793 said:
I posted this yesterday
https://toot.pikopublish.ing/notes/9wkve0zxbwjadsqh

I love that classic rock band, "Clasc Roltiock".

Yeah I think they split up after that, the profits weren’t enough once they were distributed throughout all the band – they obviously didn’t learn the lesson from Earth Wind and Fire

AudioGus

@NeuM said:
A lot of exciting developments in both still images and generated video recently. Nice examples, @AudioGus.

These same folks have a video model coming out soon too.