AI & LLM
| my first so far are gemma 4 2b and qwen 3.5 4n and ideogram 4 as image model
| While I do understand the utility of LLMS when used to assist with coding it still eludes me as to what GenAI is genuinely good for aside from porn and cheap adverts for scams.
| If we are talking images I have things to say.
I've had good success with propaganda posters and logos.
For big images regional prompting works much better than single prompts.
Also specific gooning material.
| >>1101361 personally it always felt like trying to direct an alien to make something it only had encyclopedic knowledge of yet no actual experience with the thing it's trying to recreate.
also it's virtually incapable of doing symmetry even with the newest models, it still can't get lines right in accordance to linear perspective. that's why most people mainly do gens with characters in focus but not the environment (also because goon obviously).
| >>1101363 because most models don't have any concept of geometry or physical world. This is less of an issue with models that have visual perception.
| >>1101370 i have yet to see a single one that is capable of any form of "perception", it always comes off ass exascale photobashing (which it basically).
| >>1101374 well both still get the same kinds of details wrong so i guesd that's why i didn't notice.
Refresh
| share you fav llm models and image gen