But they conveniently leave out that it costs money to do anything with AI. It's more like "open to anyone with a credit card." The vast majority of people don't have computers powerful enough to run generative AI models locally, and even then, server farms with a billion GPUs will always produce better results

This means that people have to rely on corporate platforms where you buy tokens that you use to get pulls at the various AI slop slot machines, hoping you get something decent. The mechanics more closely resemble a gacha game than any kind of artistic process

By contrast, learning how to draw, animate or make 3D models costs nothing. There's free tutorials and tools everywhere, and you can also just pirate commercial ones if you want

(page 2) 39 comments

sorted by: hot top controversial new old

[–] XxFemboy_Stalin_420_69xX@hexbear.net 6 points 3 days ago (1 children)

> caring about AI bros' opinions on literally anything

couldnt be me

load more comments (1 replies)

[–] RuthBaderGonesburg@hexbear.net 6 points 3 days ago (2 children)

Idk I just made some invitations for a party my mom is throwing. The theme is kind of niche so I used AI to generate the background. Only took 3 tries to get something that she loved, which was within the free daily limits.

load more comments (2 replies)

[–] sudo_halt@lemmygrad.ml 5 points 3 days ago (1 children)

You can run stable diffusion on fucking android phones. Your thesis is invalid.

[–] doublepepperoni@hexbear.net 2 points 3 days ago* (last edited 3 days ago) (2 children)

Really? I might look into it, then. I once tried generating random pictures with Bing (?) since Microsoft handed out some free points, and while it was fun to mess around with, I hated how much it felt like I was playing some freemium mobile game. I also just hate this sort of subscription/microtransaction-based cloud computing SAAS bullshit in general

I might actually enjoy generating stuff on my own hardware

[–] sudo_halt@lemmygrad.ml 4 points 3 days ago* (last edited 3 days ago)

To actually use it properly, install ComfyUI and get in deep. The whole "actually using SD to generate what you want consistently" is a form of art on it's own, it's kind of like learning Krita.

Because generating random bullshit ala ChatGPT is easy. Getting anatomy, object placement, logical consistency in the image and also overcoming model biases or introducing new biases is some complicated shit. For example, there are models that can generate depth layers, models that can dictate object placement, and then since SD output is smol you need to wire it into an upscaler model, etc.

SDWebUI is much simpler, but also much more limited. Same as Android apps that run SD locally, really nothing matches the level of control that ComfyUI has.

All you need is any recent Nvidia GPU with 4GB of VRAM although less VRAM was reported to work. For AMD, since AMD hates GPGPU apparently, only their top cards support ROCm. As for Apple and Intel, I do not know.

If you are stuck with AMD and can't use ROCm, just use KoboldCPP. It will be significantly shittier in every way but atleast it runs.

[–] Le_Wokisme@hexbear.net 2 points 3 days ago

never looked into it beyond "lol minions in heironymous bosch" but i see people with 10+ images of the same subject so i figure there's some ways to get consistent-ish results out of the things that the free thing isn't configured for

[–] la_tasalana_intissari_mata@hexbear.net 4 points 3 days ago (1 children)

For the smarties saying you can run it on a phone, a shitfuck scratched phone costs like 20 dollars at minimum and you're going for hundreds if you want an actually usable phone, a pencil and some sheets of paper cost way less than that

[–] Le_Wokisme@hexbear.net 5 points 3 days ago (1 children)

dirt and a stick are free but people want to make something that looks like their intent and a ton of us are not capable of that. i don't think the people writing ai prompts are meaningfully making anything but they're getting output that looks like something and if they're lucky it has the correct number of fingers.

load more comments (1 replies)

[–] americanzgenozida@thelemmy.club 0 points 2 days ago* (last edited 2 days ago)

I made a video clip using real images passed through stable diffusion on a rented gpu. It costed me 3 dollars.

I like the glitches and artefacts though, Im not trying to have a realistic result. It's more a support for the music.

In general i think technology is a theater of the class war, things we have to appropriate

load more comments