Hassaku aims to be a anime model with a bright and distinct anime style.
My Discord for everything related to anime models and art.
If you'd like to support my work, you can do so at: SubscribeStar, every bit helps and is truly appreciated!
____________________________________________________________
Supporters:
Thanks to my supporters tamashiicolle, pttcot, SETI and Kodokuna
____________________________________________________________
Using the model:
The model was trained primarily on images with minimal disruptive elements, such as floating text, logos, speech bubbles, or signatures. If any of these elements appear in an image, please include "signature" in your negative prompt.
Danbooru Metadata and franchise tags are not used in training, so please avoid using them in prompts. Metadata such as "highres" and franchise tags like "re:zero kara hajimeru isekai seikatsu" are not used in training.
A CFG scale between 3 and 7 is recommended. For best results, use CFG 6 with the Euler a sampler.
Refer to the example images, to understand, how the model should be used.
The model is designed to be simple and straightforward to prompt. Use the following tag order:
Number of characters
Character Tags
All remaining tags
Example: 1girl, rem \(re:zero\), standing, masterpiece, upper body
Here are some resolution options for SDXL:
1536 x 640
1344 x 768
1216 x 832
1152 x 896
1024 x 1024
896 x 1152
832 x 1216 (most recommended)
768 x 1344
640 x 1536
Lora use:
NoobAIs Loras working for the most part better in compare to Illustrious loras
Versions 2 and 3 is less compatible to most loras trained on Illustrious or Noob, because it got further trained away of both. Style A is still the most compatible for it.
______________________________________________________
Version and License info:
Below V1 merges ANIMAGINE XL 3.0
Version V1 use Illustrious-XL & WAI-NSFW-illustrious-SDXL with additional training
V2 is trained on its own and don't include any extra merge. Base was V1, it is trained to include newer or missing characters. It was also used for further training tests
V3 is a merge of Illustrious-XL and V2.2, to fix some issues and was also trained to include newer or missing characters
All Models using the Fair AI Public License 1.0-SD license.
Description
Be aware on 0.2: Sometimes some concepts on some resolutions looking really basic and flat. Model remains propably long as beta version. Concentrate now mostly to improve capability to promt certain characters. My model can do genshin and blue archive characters at some extend, but have a hard time on others.
FAQ
Comments (13)
Be aware on 0.2: Sometimes some concepts on some resolutions looking really basic and flat. Model remains propably long as beta version. Concentrate now mostly to improve capability to promt certain characters. My model can do genshin and blue archive characters at some extend, but have a hard time on others.
Thanks for the update!
0.2M version implements Animagine into HassakuXL in a way that cancels eachother weaknesses out.
- Animagine gives much needed style coherency to hassaku
- hassaku gives much more nsfw stability, bit better lighting and plasticity
But masterpiece, best quality in positiv and worst quality, low quality, normal quality, signature, blurry, in negative are mandatory. With additional nsfw in negative, if you want more precise sfw art.
you aint gonna update and base your model on "IllustriousXL" ? I think your model has quality, so if you do a base of "IllustriousXL", and with your dataset finetune, I think it's gonna be a good model.
Have been using Hassaku 1.5 v1.3 for... since its initial release I think, and still do today, I love it. How much do you feel the XL version is getting close to your personal satisfaction it reached for the 1.5 v1.3 version? I'm getting out of space for a new checkpoint so I can't stack every version. Would love to try it when it has reached your level of satisfaction for 1.5 v1.3!
Hard to say, because i'm basicly never satisfied, there is always something i want to improve. But I prefer the latest version over my 1.5 now. Thanks of the team creating animagine, it got the style i wanted from the beginning.
@Ikena OK thanks for the info!
Been following this project since Hassaku (sd 1.5) and for a long time this has been my favorite sdxl model by far and it keeps getting better, love the effort made by Ikena who keeps improving it <3
I think this is my favorite anime model. Thank you for making it!
One question though, I figure you recommend Euler a since all your samples are on it right? I noticed DPM++ SDE Karras and co. tend to make harsher lines, unlike the softer ones you usually see with Euler a. Which is precisely why I like this model.
I prefer 0.2 to 0.2M
0.2 is more of a 3D anime style with higher quality and follows prompts better. Occasionally it turns out a low quality dud. Do not use "masterpiece, worst quality" etc. prompts with this.
0.2M is consistent in a more drawn style. Must use "masterpiece, worst quality" prompts which can result in less prompt cohesion.
Yeah, the public consens falled to animagine and i go with the flow, training on 0.2m further
@Ikena Did it? Or is it just more downloaded because it's the first in the list.
I dunno, I really think 0.2 is much better.
Just wanted to say thanks Ikena for the quality stuff. I've been using your models since grapefruit now and 0.2M convinced me to finally start training loras for SDXL instead. Thanks for the content, and I'm excited to see where you push the model to next!