Gal 3DCG Style Lora Collection (WIP)

The 3DCG style used across multiple series. Many styles will include character tags (you may need to make additional character loras using certain lesser-known listed character tags to get more accuracy).

⚡Best Practice for Prompting:

masterpiece, best quality, @styletag, character tags. natural language prompting. booru tags (e.g. 1girl, solo, etc...)

🖼️ Preferred Resolutions:

832 x 1216 or 1216 x 832

Or Higher Res:

1024 x 1536 or 1536 x 1024

Tip: I like to do 832 x 1216, then use img2img to upscale by 1.25 with 0.25-0.3 denoising strength at 30-50 steps.

For a more 3D Look:

🎨 Situational Helpful Prompts:

ultra-HD, 3d, rendered, colorful, depth of field, detailed background, blurry foreground, blurry background, bokeh, volumetric lighting, realistic, photorealistic, high resolution, cinematic lighting, sharp focus

⛔ Situational Good Negative Prompts:

cartoon, crayon, graphite, abstract, anime coloring, manga, sketch

You can use any Anima checkpoint model with this lora, though more realistic or 3DCG base checkpoint models may produce better results depending.

Specific character tags were trained on some of these style loras, and I included the character tags in .txts. If some characters don't look as accurate as you'd like, you may need to make an additional character lora on top to get it more accurate (e.g. 'sue storm', 'widowmaker (overwatch)', etc.). These loras are focused on style.

When generating characters however, ensure you use this format: widowmaker \(overwatch\) . It should include the two \