This is my made-up *punk genre, and I love it :D
I recommend: revAnimated, weight of LoRA about 0.5, and easynegative TI https://civarchive.com/models/7808/easynegative in negatives.
Enjoy!
I offer all my content at no cost to you. Should you wish to show your appreciation and support, kindly consider buying me coffee via the following link: https://www.buymeacoffee.com/konyconiN. Your generosity is greatly appreciated!
Description
FAQ
Comments (19)
mmm ... seems can make it work. Same model, same config, dont have that totem look there.
Hello. I've just noticed, that the effect of this LoRA is negligible, when I don't use easynegative TI ( https://civitai.com/models/7808/easynegative ) in negative prompt. Please try it and let me know if it worked for you.
@konyconi same thing. no such look. with easynegative in neg prompt.
@decten76 Could you send me the prompt?
@konyconi "<lora:totempunkai:.55>, totempunkai siege engine". the lora named "totempunkai".
@decten76 OK, I don't know.
- if the file is placed in models/lora and named totempunkai.safetensors (i.e. you renamed it, because civitai downloads it as "totempunkai_v10.safetensors"), then I cannot see any reason why it does not work. If you somehow find out, please let me know.
@konyconi have u tested it on other models? have u tested on other machines? guess in IT, stuff needs to b tested in different env before releasing. What a shame, i like the look of it.
just in case, did try other loras of ur, ie clockwordspider. it works fine.
@decten76 Yep, tested on four models, got very consistent results, it triggered every time (check the gallery below). Before that, I generated thousands of pictures, scrapped like dozen versions before I was happy with the result.
Here are some ideas how to find the problem:
1] Does it produce the same picture for prompt with the lora and without the lora? For example "<lora:totempunkai:.55>, totempunkai steam engine" and just "totempunkai steam engine" (for the same seed).
2] Do you get any error messages in the console?
3] Can you see the LoRA when you click "show extra networks" and then "Lora" tab?
I've just checked on another computer.. worked OK. I hope, we find the solution.
I didn't have much problem getting it to work with REVanimated / SDE++ karas but the effect was a bit seed reliant, usually the effect was quite good and sometimes not. I'm hoping to have time to do a proper review of the v2 today and i'll be testing on a couple of models. One thing with konys LORAs is they use descriptive captions so some words have little effect
In @Mytee case I can see there's a comma directly after the Lora in the prompt, apparently this will cause the Lora not to work (though ive only read that anecdotally)
@konyconi got it work already. Followed the instruction from SteveWarner. thx.
Woah! This style is freaking awesome! Great job!
by any chance, have u made it work? or u just blessed it via the look only?
@decten76 Just responding to the posted imagery. Loading A1111 now to give it a try.
Okay. So does this work? Yes. It does. It adds a "totempunk" flair to your images. But the prompts provided in the samples are deceptively simple and if you try using a one or two word description along with the Lora you will not get good results. You still have to do some prompt engineering, but once you do, the results are pretty awesome.
Here's the setup that worked for me:
Front End: Automatic 1111
Model: revAnimated 1.1
Clip Skip: 2
VAE: Blessed2
Restore Faces: Off
Highres Fix: On
Upscaler: 4x_foolhardy_Remarci
Base Resolution: 512x512
High Res Upscale: 1.5 (768x768)
HighRes Denoising Strength: 0.35
Using this combo, I was able to get consistently good images. The prompts typically go like this:
<lora:totempunkai_v10:0.65>
Subject
Setting
Lighting
Standard Additions (best quality, detailed, etc.)
Negative prompts are the standard fare:
nude, easynegative, ugly, bad quality, worst quality, bad-artist
Note that easynegative and bad-artist are both TIs.
Also note that the Lora is weighted to about .65. At 1.0, your images will come out scorched. So keep the Lora weight fairly low as this thing is overbaked.
My early tests used the word "native" in front of things and ultimately replaced that with "totem" which produced fairly consistent images. If you're not getting what you want, try adding the word "totem" to your prompt. Just note that it will anthropomorphize most things and is prone to producing "werecreatures."
I tried using this with other VAEs and the results were bad. I can't explain that. I can only say that without the Blessed2 VAE, I got really bad results.
The RevAnimated model obviously contributes quite a lot to the overall aesthetic as does the VAE. If you use this with another model, YMMV.
@SteveWarner thanks for the detailed response. Following your instruction. Yes, it works. ;-)
Like the idea
One of the most original, coolest things I've seen posted to Civitai. Awesome.
That's brilliant, I wish we had more models like that.














