LEOSAM AiArt Turbo + LCM

This model is an accelerated version of the AIArt illustration large model combined with SDXL Turbo and LCM technology. With the Eular a sampler, images can be generated in 6-8 steps, which is three times the speed of the original sdxl version.

The effects of this model are optimized for the Eular A sampler, and only the Eular A sampler is recommended for drawing.

After multiple rounds of testing, the latest version obtained the optimal fusion ratio of the SDXL Turbo and SDXL LCM models. The current test results are the same 8-step chart. The results are: Turbo+LCM dual fusion > Turbo single fusion > LCM single fusion.

The 8-step image quality of the Turbo+LCM dual-fusion version is already very close to the original AIART model!

The Turbo+LCM dual-fusion version is consistent with the original AIART version in terms of memory usage, so if the memory is sufficient, it is recommended to magnify the straight image by 1.5 times (same 6-8 steps). The accelerated model can use a time similar to the original XL model to produce a 1024 resolution image in a time similar to the original XL model to achieve +1.5 times magnification of the 1024 resolution image.

Recommended raw chart parameters for this model: Sampler: Eular

A (Important!

The model is specially adapted for Eular a, other samplers are not effective) Number of sampling steps: 8 steps (6 to 8 steps are acceptable) CFG scale: 2 (

important

!

For the LCM model, CFG scale is recommended 1.5 to 2.5)

Magnification algorithm: ESRGAN 4x (other amplification algorithms are also available, not required, please make sure the GPU has sufficient video memory) Magnification factor:

1.5 times

Number of amplification steps: 8 steps amplification noise reduction factor: 0.3