- This topic has 12 replies, 4 voices, and was last updated 1 year, 4 months ago by
.
-
Posts
-
Rather than to continue to hijack Diviniren's art thread I'll start a new one. As mentioned I'm making a Marcel Marlier style picture book for my nieces in SD. The problem is that continuity in SD is next to impossible as it does it's own thing with each render. I'll see what happens.
I was thinking about that actually....
If I understand correctly ( can't seem to get SD working on my pc) The outcome of every render is somehow "Random". Do not know if it is possible, but say I want to make a movie with SD, how to keep the main chars from " changing" constantly? Or is that for now not possibleWell yes and no. If you make a pic in txt2img it will be completely random within the prompting guidelines, but not necessarily even that. But if you reuse the seed and use "extra", not to be confused with extras, and play with the "extra" denoising strength SD will maintain most of the initial image's characteristics/layout, depending on how much you adjust it away or nearer to the prompts. More extra denoising will give SD more freedom to ignore the prompts and change the image and less will keep it closer to them. Christ only know why SD would use a similar name (extra/s) for two different functions.
Apart from that mess I usually save out an image and fix problems in a photo prog, and then bring it into img2img to further fix things like fingers/eyes/face restore with inpainting. Then send it to "extras" to enlarge and/or adjust how much smoothing/diffuse glow that extras seems to do. Those saying that SD is a one-click "make art" button need their fucking heads read. All of the images above involved maybe 100 initial renders, perhaps a few less adjustments within extras and then multiple tweakings in and out of SD. If you're having a really lucky day you might be able to get away with less, but it ain't easy.
Always end on a song LOL..........j'aime ce genre de rendu à la "limite" du dessin. les images partagées sont vraiment super sympa, j'aime beaucoup le graphisme.
SD 101, Stable Diffusion is an AI latent text-to-image diffusion model capable of generating photo-realistic images given any text input. It has no similarities to 3D object based progs like DS or Poser or Blender. SD just makes images based on text (words) prompts fed into the program.
@Frank21 OHHHH ok, i understand now. thats pure AI creations. Thanks a lot for the explain.
Even it is really great looking, i don't like this process 😉
I would like be able to obtain such kind of render, but with my own Daz images.Each program has its own advantages and drawbacks. With SD you can't really continue from where you left off. But with SD the characters can be hyper-real and the clothes look very realistic compared to DS. "Horses for courses", ou ce qui convient à une personne ne convient pas forcément à une autre.
- You must be logged in to reply to this topic.
