Fixed the feeling of overcookness of the LoRA, it's now a bit less saturated and should work a bit better with other Style LoRAs. Also, fixed the ears a little bit, but they're a thin chance the tail might have a pointy tip. Can't have it all, huh ?
⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻
I started using Pony Diffusion recently, and was a bit bummed Smolder wasn't integrated on the model, so I did my best to port it to this wonderful checkpoint !
Almost every major issue has being fixed, the tail is way more present and unglitched !
Warning : I found the best strength to be applied to the LoRA is 0.85, but anything between 0.6 and 1 shall be good too.
⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻
Wh-what ? A new version ? Yeah, sort of. I noticed a bit too much problems with the previous 1.0.1 version. I decided to try and fix them, the result ? I fixed it. But as always, problems remains. But here's the new version !
Better color
Slightly more stable tail
Fixed an issue where the sweater would stay on no matter what
Better tagging
Few issue remains, sometime the tail is still a bit ??????, or sometime it makes two of them. And the issue I've come against the most, is that sometime Smolder has 4 eyebrows. Pretty annoying, yeah. This version seems a little less versatile in terms of checkpoints.
So if you struggle with this version, if it doesn't produce quality images, I recommend you use the old 1.0.1 version, which has problems too, but is a bit more versatile. I hope you'll like it nonetheless, feel free to posts as much images as you want ! Thanks !
⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻
This will most probably be the last version, I cannot do any better than that, with my methods of training. I hope you'll like it, don't hesitate to show me your generations ! I'd love to see your arts !
What's new ?
Slightly better stability
More sample images
Better tagging system
Slightly better tail implementation
I don't think I'll do another LoRA for a long time, it was painful to make !..
Small issue :
It doesn't work well with other concept/action LoRA.
It has a rough artstyle, with some checkpoints.
Like many other models, I'd recommend using this LoRA with Hires Fix. I personally use any upscaler at 1,65.
⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻⸻
But I hope you can still enjoy it !
I'd recommend a 0.85 to 1 weight. Weight makes a huge difference.
Triggers are Smouldie mainly, then the rest ; 1girl, blue skin, masterpiece, best quality, short hair, (tail)
As I already said, this LoRA isn't perfect. If someone is skilled enough to produce a better Smolder LoRA, please, make a better version ! I'd be so ???? happy.
Artist's page :
https://www.patreon.com/Kanel_Art/posts
(Don't comment to ask me to fix stuff, this is the best I could do, I'm not good at this..)
Update on Pony, less "overcooked" feeling. More consistant with the ears.
1. Modèle partagé uniquement à l'apprentissage et au partage. Droits d'auteur et interprétation finale réservés à l'auteur original.
2. Auteur souhaitant revendiquer le modèle : Contactez officiellement SeaArt AI pour l'authentification. Nous protégeons les droits de chaque auteur. Cliquer pour revendiquer
