Most image benchmarks test whether models can place objects and attributes.
I wanted to know which ones offer photographic control.
So I built the Filmmaker’s Benchmark to evaluate how film stock, lighting, and scan-treatment prompts change color, contrast, and texture.
My ChatGPT account got hit by the auto-ban wave, and when they reinstated it, it was only on the free tier. I had a pro subscription with most of the month left. Can I get the credits I paid for? @thsottiaux@sama
@reach_vb My account got reinstated, but only on the free tier. I was a pro subscriber and had most of the month left. Can I get the credits I paid for?
@LexnLin I've been testing it out in ComfyUI, and so far I still prefer Z image (base and turbo). Ideogram's prompt adherence is very good, but it has that AI texture I just can't get rid of. Meanwhile Z image can actually generate photorealistic stuff
@NousResearch The desktop app is fantastic. Really good information hierarchy and UX. I’ve been bouncing between few different apps and dashboards for Hermes and I think this is gonna be the one that finally sticks
@somewheresy@cairoasmith “If everybody is an ‘artist’ then
nobody’s an artist. “
I remember how a decade ago this very same crowd was arguing that drawing is more about personal style and self expression in order to justify relaxed technical requirements in American art schools. How ironic