CHECK THE "ABOUT THIS VERSION" sections.
Just as before, this is a TensorArt online trained model.
This time, I applied a very rigorous labeling process to all the training images—which were fewer in number than those used for the Kennauri Style LoRA—resulting in a more focused and consistent style.
Furthermore, thanks to this tagging, the LoRA does not apply—and did not do so in any tests—hairstyles, clothing, gradients, or jewelry that are not strictly included in the prompt.
The recommended weight is 0.8.
Recommended Samplers: Euler a/DPM++ 2M Karras (I didn't test the others) and Normal or Karras Scheduler.
Just as before, do not upload my LoRAs to other platforms without my permission, and do not use my characters (shown in the examples) unless you give me credits.
Description
Technically, the same style that we had for Illustrious.
Always remember to add the quality tags to secure the best results: masterpiece, best quality, score_7, absurdres,
If you don't know much about Anima, you can use the default as Negative Prompt.
Since Anima doesn't work with Adetailer, it is recommended to use the Hi-Res Fix.
I recommend 4x-UltraSharp or Latent (bicubic antialiased) with 10-15 steps and 0.35 denoising strength for far/wide shots and full body.
For closer shots, face focus, half body, use R-ESRGAN 4x+ Anime6B with 15 steps and 0.20 denoising strength.
DO NOT use the Upscaler on 2x. It will make the images take a noisy/granulated texture. Use it on 1.5x.
Do not use the LoRA on 1.0 strength. Use it on 0.8 or 0.9