Large New Dataset 220k AI Art Text to Image Prompts
github.comInteresting to see, lots of people using modifiers like detailed, artstation, beautiful, ghibli, octane. Fairly sophisticated stuff people are using but i dont think the understanding capabilities are there that much yet for art generators. Dalle3 probably the most factually correct, midjourney and ebank.nz look nice
Seeing incremental gains like negative prompting, but then all of a sudden theres huge progress like latent consistency models, stable diffusion turbo/dalle etc.
Pretty exciting time, i think UX having a big moment too :)