Caradhras~3dcg style | Illustrious - Experimental

Caradhras~3dcg style | Illustrious

LORA
Original


Updated:

After spending countless hours working on this, I can definitely say this is the best results I've gotten so far, despite my best efforts, Illustrious has an issue with 3D faces and eyes, and it pains me that I can't do anything about it, no matter how many times I rework the dataset and the images, no matter how many times I go over my settings while training, this is by far the best one I got and I'd still recommend either going for more of a cowboy shot angle or portrait, you'll get way more details. I'm tired, I'm frustrated and I can't be bothered to train the 20th lora for a 1~2% increase in face quality, so I'm releasing this as is and I hope you guys enjoy it. If you want to support me, you can either donate, like or follow for more updates, either one helps! Thank you!

Recommended settings:

Strength 0.9, lower for more flexibility. Same goes for your cfg, anything below 4.0 is perfect or recommended. Also try not to use too many LoRAs together as it can diminish the quality of your image fairly quickly, or a high cfg will also do that.

Here's some tips, any lighting tag will improve the results by a mile, if you check my prompts used in those images, I kept it as basic as possible so y'all could see the potential. Also, tags such as realistic, photorealistic, 3d, blender_(medium), cinematic, raytracing, chromatic aberration, film grain, etc, will also influence quite a bit the results, up to you.

Thank you again for using my LoRA, you can support me through either donating Buzz, or you give a thumbs up, follow me or you can also use one of the links provided beneath my profile, anything helps really.

Version Detail

Illustrious
<p>Another version I've been working on, unfortunately Illustrious just hate 3D style, specially based on 3D videos similar to styles like Rigid3d, or Nyl. I've been struggling to get a very well made lora that doesn't need to rely on tags such as; Photorealistic, 3d, blender (medium), realistic, etc. I wanted to make a lora where you can simply plug it and play with it and the style will generate easily and spare you the tokens for other things, this version came very close to it but you can still catch some facial bugs such as teeth and bad eyes, unfortunately there's nothing I can do to help with that, that's just IL and 3D hating each other.</p>

Project Permissions

    Use Permissions

  • Use in TENSOR Online

  • As a online training base model on TENSOR

  • Use without crediting me

  • Share merges of this model

  • Use different permissions on merges

    Commercial Use

  • Sell generated contents

  • Use on generation services

  • Sell this model or merges

Related Posts