[–] 4 points 11 months ago

I would personally prefer to not have AI Generated photo realistic content that can easily be mistaken for real photo in this community.

At the same time its very nature makes it really hard to moderate and control, and can easily spiral into witch hunts or limiting the content from only an handful of trusted sources, which would kill the already limited contributions outside of the ones from the wonderful anon.

The only solution I can think of is adding a vague rule, as someone already posted, asking to avoid it and be lenient on the casual trasgressors, asking, educating and warning first. Not straight out banning or demeaning.

At the same time I feel there should be low tolerance towards the amount of vulgar, harsh and honestly disheartening comments that tend to flood discussions even vaguely related to AI on lemmy. There's a really vocal group of people that floods any thread that could potentially be AI or about AI.

I get the reaction, but I think it has no place here: this has always been a wholesome, kind, community.

About artsy AI (drawings, comics, fake paintings, digital pieces etc) I don't really have a strong opinion, mostly because I don't follow this community for the art but mostly for the photos of real owls: I just think it should at least be properly tagged.

  • source
  • [–] 6 points 1 year ago (1 child)

    I don't have direct experience with RooCode and Cline, but I would be mighty surprised if they work with lesser models of even the old Qwen2-Coder 32B - and even that was mostly misses. I never tried the Qwen3 coder but I assume it is not drastically different.

    Those small models are at most useful for some kind of smarter autocomplete, not to run a full tools framework.

    BTW you could check out Aider too for a different approach, and they have a lot of benchmarks that can help you get an idea about what's needed.

  • source
  • [–] 2 points 1 year ago*

    I don't have them: I generated a new one modifying the prompt

    Is this what you meant? If you want it in other styles I can try them out, but it will take some time

    EDIT: If it was because I said it was straightforward to gen them in whatever style, it is because of the dataset Chroma used for training: I would be incredibly surprised if centaurs aren't in there

    Sorry about the confusion

  • source
  • parent
  • context
  • [–] 4 points 1 year ago (2 children)

    Using Chroma (v50 annealed GGUF Q4) it was pretty straightforward to get a centaur in whichever style, so I tried to make something weirder, making the head a completely different style from the rest of the image.

    It was kind of a PITA to get something somewhat coherent and not messed up, but I think with enough prompt massaging it could output something really nice, but I don't have the hardware to experiment too much: it just takes too long to gen.
    Still, it was fun.

  • source
  • view more: next ›