5

Can the Dev put this question to rest? How is the text processed now for t2i? Is Automatic1111 still in use? Stepping and blending don't seem to work after the Model change or at least not predictably, but weights do work predictably. I can see where the properties are stripped, an DSL evaluation runs, and the word filter, but what is processing the weights now? A frontend or are the handled by the CLiP and T5?

[-] wholeai@lemmy.world 1 points 5 months ago

If I'm reading this right, you probably want to override the CCS for the bar. I've seen this on a few pages. Best thing to do is find a page that does this see what their doing to make it happen. Copy the JS from the page and give it to any big name chatbot, it should be able to find where this happens, explain how it works and how to incorporate it into your gen.

[-] wholeai@lemmy.world 1 points 7 months ago

Letting the days go by

[-] wholeai@lemmy.world 1 points 8 months ago

Pretty sure you could use this gen to give the prompt color names instead of hex code hex to color name generator

[-] wholeai@lemmy.world 1 points 8 months ago

Oh damn that formatting.. here the screencap

[-] wholeai@lemmy.world 1 points 8 months ago

Sub lists can be a pain. I'm sure there's a better way to solve it this problem, but I think this gets the output your looking for. The aspirinPeople and ancientPeople work effectively the same, just to show the variation. If you called ancientPeople without selectOne it iterates through the list.

output [r = ruin.selectOne]

ruin [Ancient] [Aspiring]

Ancient Ancient graves portected by [ancientPeople.selectOne] Aspiring You come across an aspiring [aspiringPeople] ancientPeople mad dog sad jim

aspiringPeople {ninja turtle|greedy cleric}

wholeai

joined 8 months ago