Yeah! It’s good to publicly and independently document their strengths and weaknesses because all the big companies are financially incentivized to promote their models.

I found it very interesting that you could easily fool the model by simply inserting praise for the paper inline.

Great! Thanks for doing the test and being open about the results. Looking forward to a paper on this.

Did you test this with Opus?

I have reached out to folks yes :)

Time to sleep

What use do the dead have for gold anyways

Gosh a dynamic pose in the style of a movie poster showing off the wizard casting a particular spell would be cool. (Note: I would only want to make this a single character)

Hi folks, thanks for the recommendations and self-recommendations. I’ll do a write up of the subject, the style and other ideas in another comment tomorrow! Feel free to express interest in the meantime.

70s, 80s pulp fantasy novel orrrr a movie poster from that era

Yes, I’ll update this thread with requirements tomorrow!

Man a full movie style pose would be nice!

Check out my other replies in this thread :) I’ll also add a pinned comment with more details after I get home

(Please don’t draw anything yet) Little gnome wizard dude. Youngish, no beard, spunky, red hair, has a robe with a ton of pockets to easily grab spell components.

I would like a 70s-80s pulpy fantasy zine vibe, or like a movie poster vibe.

Perfect :)

True, so the artifacts are similar to image compression artifacts. It’s a good thing they trained to make it robust against noise addition!

Very cool, have you noticed any image artifacts? It really doesn’t seem too bad!

I’m considering switching to base Chromium.