[maalber]

Very interesting - definitely looks better than what i mostly get from DALL-E Can you share some technical details on how you did it? Did you train your own model? Or fine-tune an existing?

The option to polish it in figma is interesting. I have made a figma plugin that lets you have 100 people vote on their preference between two figma frames to select the better looking one. Maybe there would be some potential for a collaboration. I imagine your model generating two (or more) logo variations, and then you can immediately find the preferred one through the voting. Feel free to check it out for free at https://app.rapidata.io/compare or https://www.figma.com/community/plugin/1418578607167949732/r...

[Abod]

Yea, I've tried multiple models, didn't like most of the results, so had to fine-tune my own.

That such a noval idea, I might use at some point, thanks for sharing!

[maalber]

Yes, I also feel like most of the off-the-shelf models struggle with generating nice and natural looking logos, especially something that resembles vector graphics. Since you have been fine-tuning the models yourself, I really think the tool I mentioned could be useful for you. The one I linked just obviously only lets you queue up one pair at a time, however, we also have an API that lets you get responses on 100s or 1000s of pairs quickly. Please feel free to reach out if you want to try it out. My email is on my profile.

[Abod]

Really appricate your offer

Will certainly need at some point, btw does it work with a set of 3 instead of 2?

[maalber]

The version we currently have open is only 2, but we are soon going to release a version that is more general ranking which can do 3 and more options. Hit me up and I can keep you updated when it is available!