ฝัง
- เผยแพร่เมื่อ 27 ส.ค. 2024
- It works like you’re seeing me do it on screen : you go to their website, rate a couple hundred images, and then once you’ve finished you can go to the discord interface and use the -p command to specify that you want to use the personalized function.
Now I’m going to explore what this all means for us.
Become an insight collector with Napkin, the perfect thought-capturing app :
napkin.one/?vi...
really loved your breakdown of the advantages of hand drawing thumbnails!
Thank you so much I appreciate that !
Why not both... thumbnail drawing by hand to do the "thinking", have AI fill it to give it as a first impression to the client, gather feedback, adapt the thumbnail drawing if necessary, then work on the final piece.
I've thought about thumbnails and AI for a while and I think the the relationship is somewhat backwards. I'd love to be able to let AI take my thumbnails and render them. But instead its just rendering fairly boring compositions. This is a place where I think AI and artists could benefit from each other. Thumbnails are where a lot of the artistry can occur so even though you may be leaving the rendering to the AI, it still could be considered art that you made. You're just leaving the time consuming "grunt work" to the AI. Instead they're trying to go all the way to replacing artists and that doesn't appear to be working.
That's a lightbulb moment for me, I'd never thought about it that way..
Thanks !!
As a probability engine, AI really struggles with perfection, and it will take a long time to this not be the case (maybe for ever). It's in the minor things as you said. A real devil in the details. That's why it is used in programming only when supervised. Some times correcting it's errors is more time consuming than implementing the feature yourself from the ground up, as these errors are made so masterfully hidden. That's the reason that it's not being seriously used in the health industry in large scale, or in law firms to right important documents. The devil in the details can kill a company if it is hidden in the wrong place, and there will be no one to blame. So a human must read, correct and sign below.
Now. Think for a moment. Which endeavor is more perfectionist than art creation? How much you must describe your vision in a prompt to se it rendered as you envisioned it in the screen?
An infinite prompt I would say. Than why not take your pencil and draw it? It will be easier. I assure you.
If AI is capable to reproduce your vision perfectly, is this vision really yours?
Great video. Necessary discussion. Cheers
Suno song at the end? :p
Personalization? Midjourney is leagues behind local Stable Diffusion in terms of stylization, features and useful techniques. I should try it if you want to see a bigger picture
Yes StableDiffusion is much more modulable and reusable. Actually in one of my recent videos I created a new use of the stable diffusion vae to make kind of abstract art.
But it can't be argued that midjourney gives more aesthetically pleasing results. Most of the people I know of who advertise use of AI in industry recommend Midjourney.
That's also because it's much more accessible to normal people, in particular artists.
"turn randoms into full-fledged artists" And I'm here thinking artists were considered randoms because society loves to ignore them. Have you seen the number of good artists ignored on Instagram, Facebook, X, etc? It tells a lot about what people think about artists.
(yup, I'm an artist)
On the point of comparing an AI image to a thumbnail, I hold that
A) a thumbnail is fast, seconds or minutes rather than hours. [If it is hours to make a thumbnail, that is an interesting oddity]
B) a thumbnail is not (usually) an end to itself.
This is a choice of framing of how to view AI art, as a quick stepping stone in a iterative process.
It doesn't mean that AI has value for any given artist. We have to sort of squint and assume value for this to even make sense to discuss.
This choice of framing also doesn't say anything about AI replacing thumbnails.
On the contrary, a thumbnail would make perfect sense as part of an img-to-img workflow (show me 100 loose-polish versions of this thumbnail).
So in that case, the thumbnail would come first, then the renders, then whatever other process.
I'll add, I'm beginning to doubt my own framework here, for reasons I won't go into just now.
It's by no means a perfect system.
I'm not interested in MJ (for a dozen reasons), but I REALLY like this video.
FASCINATING.
Most artists are randos though.
i guess you're saying art if you mean illustration throughout the vid.
Indeed there's a big distinction there. AI can already do illustration. The topic of this video (and channel tbh) is whether AI can step into art as well, and if so, in which ways. By knowing this, we can adapt and avoid personal tragedies
Man I hope you don't get shit on simply for being an AI focused channel, you value traditional art, and these blind raging anti-AI zealots might take away from interesting discussions to be had.
Thank you so much. It's been alright so far tbh, apart from the like ratio pattern I've noticed :)