![]() In the end I feel like consumers will decide, not nesessary creatives themselvesĬlick to expand.There's not much to discuss here, as the training process is similar to what human does while looking at a picture. Hard to say where it is going, both legally and reception wise. At some points, even hacks like in-paint or img2img may be insufficient, so manual retouching will be needed. Issues with eyes, hands, poses and even elementary logic (it can draw the belt overlapping an arm or more than 4 legs on a horse). It is trivial to produce something that looks like artwork, but for more specific tasks, you'll have to fight against the algorithm. Thats why it is, for example, exellent with faces at front angle, but messes up eyes most of the time, because look direction on source pictures was all different)Ībout how it can help you, well, as neginfinity said, will depend on the context of what you need. Neural network doesnt form understanding of, say, anatomy and perspective from examples, and only draws what is statistically more possible. (AI advocates will go on about how "it Is tOTallyY LIke hUmANs learN to DraW", no I can't agree. So no surprise many of them beel betrayed. Since artists had no idea someone will ever use their work to train neural networks, no one could opt out of this practice until it was too late. It doesn't exactly "copy" artists' work, but it had artwork (scrapped from all over Internet) compressed into itself during creation. Personally, I am mostly negative to it, partly I do not have the hardware to get decent results from standalone version of Stable Diffusion, and refuse to support online services, like Midjourney, with money. The general consensus on gamedev forums is "use it and to hell with what the artists think automation blah blah", so to get the alternate perspective you'll have to ask in multiple places ![]() ![]() Your software development skills should aid you while researching a new field. You know how some people used Daz3D to make comics and visual novel characters? It is this sort of effect.Īlso, be aware that learning how to draw is an option. If you do not invest your time and learn and just slap a basic midjourney artwork, you'll produce "fodder art" which will look cheap to people familiar with Midjourney/Stable Diffusion, despite being detailed. The instrument can refuse to cooperate.īasically, to get good results, you need to invest your time and learn. What's more, the tools are very rapidly evolving, for example, a short while ago someone introduced control net and now it is possible to pose characters. It takes time to learn how to use Stable Diffusion, for example, and how to coax neural network into producing something similar to what you would want. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |