user on Nostr: txt2img base models rated: sd15: base model is not really usable, fine tunes work ...
txt2img base models rated:
sd15: base model is not really usable, fine tunes work well, but often fail anatomy. fine tunes follow prompts very well
sdxl: base model works, but is boring, fine tunes are okish. can do some nsfw, does not follow all prompts
sd cascade: pretty much like sdxl, better quality, more complex multi stage workflows, nothing surprising
lumina: excellent sfw generator, doesnt really do nsfw
pixart: more artistic and creative style compared sdxl for example, can do some nsfw, does not follow all prompts
sd15: base model is not really usable, fine tunes work well, but often fail anatomy. fine tunes follow prompts very well
sdxl: base model works, but is boring, fine tunes are okish. can do some nsfw, does not follow all prompts
sd cascade: pretty much like sdxl, better quality, more complex multi stage workflows, nothing surprising
lumina: excellent sfw generator, doesnt really do nsfw
pixart: more artistic and creative style compared sdxl for example, can do some nsfw, does not follow all prompts