( Not my Model )

https://civitai.com/models/3042937/tinanimav1hint84steps-yuan?modelVersionId=3458928

( BF16 Version )

TinAnima_v1H_INT8_4steps_Anime Style

This model can generate large-scale images directly, eliminating the need for an upscaler.

Tested on an NVIDIA RTX 3060 (12GB VRAM) using WebUI Forge Neo: generating a 1024x1536 image takes approximately 3.8 seconds at 4 steps.

Sampler: Euler (ER SDE)

Steps: 4 or more

CFG Scale: 1 or more

Image Dimensions: 1024 x 1536

Positive Prompts: You may refer to the parameters shown in the example images. Generally, you can use natural language prompts (similar to those used for Z-Image) or standard, comma-separated English keywords typical for SD 1.5.

Negative Prompts: worst quality, low quality, score_1, score_2, score_3, artist name (optional).

You can also use the prompt parameters found in the provided example images.