👇👀 Read me! 👇👀
I'm baaaaaack! Contrary to belief I am still very much alive and always have been 😉
This a bipolar slider for the ANIMA family of models built to control the age of characters. As normal with my sliders, it is built to have as little entanglement and unintended impact as possible, which is what makes my sliders so great 😁
No activation tag. This bipolar slider LoRA supports positive and negaive input weights, with the recommended range being from -3 to 3 . Negative values introduce younger age. Higher values introduce older age.
Example prompts can be seen in the attached example images.
What other sliders do you want me to port to ANIMA, or what other brand new sliders would you like me to make? Be sure to comment below to let me know! If you encounter other issues using this slider, please let me know so I can fix it!
This slider is not available on CivitAI's web generation service due to requests from CivitAI's moderation team in the past, but you can run it on TensorArt!
~ Known issues with v0.9 ~
1.) High negative weights <=-3 can introduce slight realism
2.) Concept disentanglement can still be improved with some concepts like clothing and chest size. If you require better disentanglement - let me know!
~ Changelog ~
June 19, 2026: Initial v0.9 release on CivitAI and TensorArt
Description
Version 0.9 Initial Release
FAQ
Comments (16)
This is exactly what I've been waiting for, thank you so much!
LETS GOO! THE KING/QUEEN OF SLIDER LORAS IS HERE!
Chest size slider next :D?
@TOF_enjoyer Pretty much already trained. Now pending validation, pruning, QA, and generating example images. Might be able to squeak it out this weekend, but more likely coming sometime next week.
Absolutely astonishing! Works perfectly!
Finally I can get older women without using 6 billion tags and then ruin the image by accident because suddenly they may appear to look like 500 years old 😭
That's the idea! Let me know if you run into any issues using my slider to do this 😀
Agreed
Still, I wish someone made a lora with tags for each wrinkle tbh, like under the eyes and around the mouth
@properstyle Yeah that would be great. I absolutely love those details
@properstyle Probably not realistically possible to get it as precise enough to control exactly where the wrinkles appear with an acceptable level of quality, but I will take a look into it. Should be a good challenge even if it doesn't work out.
@Shed_The_Skin Oh I didn't mean to request, but it would be fantastic. I love designing characters and sometimes I just can't get the details as I want them because there's no tags for it. To be fair, danbooru added a few (crow's feet, frown lines, smile lines, tear throughs) but they're too recent to be in the database of any AI model.
@properstyle I started to work on a lora for those mature details but it's damn hard to find enough and good images for it. I only found 34 images so far and they don't even cover all of the details yet😭
@Lopodelas Not sure how the curating process works but I would think finding images where they DON'T overlap so that the concepts remain separate might be the hardest part, since at least in anime stuff they tend to either have all or none.
Here's some links to the tags I mentioned, hope it helps:
https://danbooru.donmai.us/wiki_pages/wrinkled_skin
https://danbooru.donmai.us/wiki_pages/crow%27s_feet
https://danbooru.donmai.us/wiki_pages/frown_lines
https://danbooru.donmai.us/wiki_pages/smile_lines
https://danbooru.donmai.us/wiki_pages/tear_troughs
https://danbooru.donmai.us/wiki_pages/wrinkled_forehead
And of course, these too:
for some reason my comfyui gets stuck on ksampler for like 3 minutes every generation only when i use this lora
@spunkymcgoo Does ComfyUI log any errors? If not, could you send an example workflow so that I can reproduce the issue?
@Shed_The_Skin
I don't see any errors in the log. Here's a workflow https://www.dropbox.com/scl/fi/9km8z8gzxqif8qm2uf022/test.json?rlkey=kfibvqqz48rwkaoptgbhclj59&st=17fuai5z&dl=0
I turned off all my extensions to test, and i noticed that if i'm not running torchcompilemodel from kjnodes, and i don't change the lora strength, it'll generate as normal after the initial wait, although it'll do it again if i change strength. Feels like it's compiling something, but i have no idea how to find out. Maybe it's a ROCm thing? (i'm on amd+linux)
Here's the log: https://www.dropbox.com/scl/fi/lu1750i05kns3jsgjacrw/comfyui.prev.log?rlkey=ythx3xmbkpbioxtszrzaatq5q&st=nnmyxamy&dl=0
@Shed_The_Skin i think it might be a linux kernel regression actually? people are reporting issues with comfyui on the newest kernel
i guess it's not your problem