v3 (Flux Dev)
Woah. Flux rips hard when training.
Best model was at 3.5k steps.
Keywords (Optional): eyes crossed, tongue out, ahegao face
Strength: 0.8 - 1.0
Edit: Apparently there was a bug in the training code that produces slight artifacts in the resulting images. I will probably update this model as soon as I get the time to train it again. Sorry about that.
v2 (Pony)
It's back. The myth. The meme. The ahegao.
This was once again a lot more work than I anticipated, but I got to try out some new techniques on this one, so I hope you like it.
New Pony Version
Much improved Dataset
New Captioning
DoRA Training
Embedded embeddings (for hopefully more flexibility in models)
v1 (SDXL)
The model is super flexible. Trained at multiple resolution around 1024px.
Trained on over 300 high quality images, so I had to leave my computer running for two days straight, but I think the results are worth it.
The keywords can be used pretty independently, except maybe "drool" since there weren't as many training images for it.
After a few requests I've added the training data for anyone who wants to create an update or a model for SD1.5 since I don't really have the time/motivation anymore :)
Also, I think I may have started a trend with those supermarket images 馃槄
Description
FAQ
Comments (25)
So it Begins
Fantastic! That drool at the tip of the tongue 馃槢
Really good work ;-) Also thanks for using Hephaistos for your showcase. I will write a review soon :-)
Impressive model. Where did you find the xl6HEPHAISTOSSD10XLSFW_v10 checkpoint? I can't find it on Google
That looks like this - handily, I saw it just a few minutes ago!
https://civitai.com/models/119279/xl6-hephaistos-sd-10xl-sfwandnsfw
Can I use it on SD1.5?
no SDXL is not compatible
please make a 1.5 version
Agreed, I don't have the pc specs to run sdxl but I can run normal sd 1.5 so a 1.5 version would be amazing
There are already 1.5 versions out there TongueDrop is a good one.
I'm amazed you got the eyes to do that. If you could train eyes looking left or right I bet you'd make thousands of people very happy! Wall eyed too would be fun. (Each pointing outward.)
Thank you for the kind words! I honestly think it's mostly a function of SDXL. I've uploaded the training data in case anybody wants to take inspiration for a "looking" model.
Unfortunately this dose not work together with my lora (a face) and Colossus XL version 2.
Can't get it to work no matter what I do :(
Hey there, I've just tested this even with new SDXL Turbo Models, and it seems to work pretty well. Are you getting an error, or is the concept just not coming through?
Do you mind sharing your training data?
Not at all. Just not sure about the legality of the dataset. Drop me your contact info if you're interested.
nvm I've uploaded the training data. There were quite a few requests for a SD1.5 model and hopefully this will help :)
@Joly0 Feel free to delete your post now ;)
@mr_chrisma聽Thanks alot for the training data
very nice. what hardware did you train it on?
I believe 980ti and 8700K at the time. But might have upgraded to 24gig card already.
@mr_chrisma聽cool. Next step would be to train on multi gpu?
would die for one of these, pony or otherwise, where it's JUST the crossed eyes and not the mouth.
Should be doable with the new version ;) (but please don't die)
This is great