♦♦Pearly Opal Anima♦♦
A checkpoint focussed on bringing a more detailed lighting and smoother shading along with more interesting backgrounds without much hassle.
♦♦Installing for local gen♦♦
Get the text encoder and VAE from the HuggingFace page.
The model is natively supported in ComfyUI. The model files go in their respective folders inside your model directory:
Pearly Opal Anima .safetensors goes in ComfyUI/models/diffusion_models
qwen_3_06b_base.safetensors goes in ComfyUI/models/text_encoders
qwen_image_vae.safetensors goes in ComfyUI/models/vae (this is the Qwen-Image VAE, you might already have it)
♦♦Settings♦♦
Settings that i use with this checkpoint
Note: I’m pretty much winging it.
1st Ksampler
Sampler: Euler
Scheduler: Simple
CFG: 5 ish
Steps: 20 - 35
Ultimate SD Upscaler
Sampler: Euler
Scheduler: Simple
CFG: 5 ish
Steps: 5 - 20
Tile size: 512 or 1024
Recommended positives:
(masterpiece, very aesthetic, hdr, depth of field, high quality, realistic lighting. masterpiece, best quality, good quality, amazing quality, extremely detailed, extremely aesthetics, intricate details, absurdres, newest, highres, score_9, score_8), Recommended negatives:
(lowres, bad quality, worst quality, long neck, bad anatomy, watermark, jpeg artifacts, ugly, poorly drawn, score_1, score_2, score_3),Additional notes:
This model performs best with more detailed prompts. Dont forget you can also use weights on prompts example: (standing:1.2),
Description
FAQ
Comments (13)
Visually, it's spectacular, but it gets too close to the characters and also almost always has one looking up or raising its head, even though it tells it otherwise in the prompt... I don't know if I'm doing something wrong, or if it's just a training thing.
Thanks for the feedback :), not really noticed the looking up thing but I'll look into it.
@Serephian As I said, it might be a problem with my prompts, but I use the exact same ones with multiple checkpoints, and I take several batches of each image, and with this one, there's always at least one character looking up or with their head raised. But not only that, when generating group images, it zooms in too close to the characters, leaving no room for everyone to appear, or they're crammed together and distorted to fit in the frame. It's a shame, because as I said, visually it's spectacular; I love the visual result it offers. Anyway, I'll keep testing it, and I'll probably just use it to generate single-character images and that's it. In any case, thank you very much for your work, as always!
yeah it does the same thing for me too, and I can't prompt it to not do that no matter what I do. It's shame because the image quality is good
An excellent checkpoint that generates truly impressive and aesthetically pleasing results. It performs best with carefully written prompts, but when you find the right words, it rewards you with gorgeous, high-quality artwork. A model that combines strong technical capabilities with wonderful artistic expression ❤️
Haven't tried it my self yet, but looking at what people make in the feed... holy, the results are amazing! 🙏
Looking forward to your gens :)
Oh ill have to try this tomorrow. I did try v1 but my wildcards was throwing up some crazy gens lol. Ill report back when iv give it a try
It's clear that some improvements have been made compared to the first version. It's a wonderful, beautiful model, and I recommend everyone give it a try. Huge thanks to its creator, Serephian
Thank you! Glad you like it :)
CHAT WE ARE SO BACK 🙌💖