Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Wow, this is pretty cool. Instead of turning a picture back into text, turn it into a unique concept expressed as variable S* that can be used in later prompts.

It's like how humans create new words for new ideas, use AI to process a visual scene and generate a unique 'word' for it that can be used in future prompts.

What would a 'dictionary' of these variables enable? AI with it's own language with orders of magnitude more words. Will a language be created that interfaces between all these image generation systems? Feels like just he beginning here..



Right on, you could see a marketplace of those mini- pretrained weights to have stuff a styles much more available in a UI like setting without needing to add style manually... Very interesting.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: