4 ms·

I’m not terribly familiar with the text to image tools, but you can provide source images as baseline, right? I’d wager that if you’re able to create a baseline

I’m not terribly familiar with the text to image tools, but you can provide source images as baseline, right? I’d wager that if you’re able to create a baseline image to feed in, your results will be better. The better the input, the better the output. It definitely feels like a situation where artists who can leverage ai will be the ones pulling ahead in the commercial sector.

keiferski 3y ago
It doesn’t really work that way. Yes, you can use images as a source, but they are more just mined for “pieces” to rearrange, not overall aesthetic effects.
- astrange 3y ago
  You can do it with ControlNet guidance for SD.
- minimaxir 3y ago
  That is not how image-to-image approaches work. ControlNet is a obvious counterexample. If you think "diffusion is just collaging", upload a control image using this space that cannot exist in the source dataset (e.g. a personal sketch) and generate your own image: https://huggingface.co/spaces/AP123/IllusionDiffusion https://huggingface.co/spaces/AP123/IllusionDiffusion
  - keiferski 3y ago
    It’s not that I think it’s purely collage, but that inputting a high-quality image doesn’t somehow lead to generating better quality output by default. The various silly images created by using keywords like “Greek sculpture” or “Mona Lisa” are an example.