models/ALF mix (IL, 8 steps generation) - v4.3 [final]

ALF mix (IL, 8 steps generation) - v4.3 [final]

8/9/2025

7:11:28 PM

Related Keywords & Tags

anime,anime image generation,base model,checkpoint,illustrious,lcm sampler,shio_n,stable diffusion,v4.3 final,v4.3 [final]

Abstract portrait of a figure with large eyes and swirling patterns in aqua and black, featuring synthetic overlay light and chromatic aberration effects.

A noir-style ink and sketch illustration showing a mysterious character with a shadowed face and wrapped arm, highlighted by a sepia spotlight and textured halftone background

$A dark silhouette of a figure standing in a dimly lit room with cobalt blue and ivory fractured light patterns illuminating behind and beneath them, featuring glitch effects and watercolor textures.$

A dark silhouette of a figure with glowing orange eyes and mouth, wearing a hoodie with a smiling sun emblem, standing with bandaged leg and heavy shadows.

Abstract silhouette of a dark figure with glowing white eyes floating above the ground in a gray and peach color scheme, surrounded by soft directional light and misty atmosphere.

Anime-style girl with blue hair and pink streak, wearing a pastel rose and blue floral dress, leaning on a wooden bridge with roses and water background.

Recommended Negative Prompts

worst quality, bad quality, low quality, (jpeg artifacts, mutated, bad anatomy, amateur, :0.7)

Recommended Parameters

samplers

Euler A, LCM, DPM++ 2M SDE, Euler a

steps

6 - 20

cfg

1 - 3

Recommended Hires (High-Resolution) Parameters

steps

Tips

Use style tags such as 'anime style' or artist/title name to help the model understand desired output.

For fastest generation, use CFG 1.0 and set lora CFG_slider on negative -0.4 value (note: negative prompt ignored at CFG 1.0).

For best coherency, use CFG 2.8 and CFG_slider lora at +0.75 weight with Euler A sampler.

Version Highlights

I doubt I can do something better than this one by mixing.

v4.3 [final]

I recommend to choose this one or v2. If you don't care about speed and want great anime results use this model + LCM sampler + 20 steps. It's simply better than any model I know in this aspect.

Recommended settings:

6+ steps (8 + 4 steps of upscaler is optimal for my use cases, feel free to experiment).

A few options of samplers:

CFG 1.8 | Euler A
CFG 2.0 | LCM - better for anime and other abstract things.
CFG 1.8 | DPM++ 2M SDE with BETA scheduler (normal doesn't work)

Use style tags such as "anime style, flat coloring" or artist/title name. It will help model to know what you want from it.

The fastest generation: use CFG 1.0 and set lora CFG_slider on negative -0.4 value. Lora listed in suggested resources. (CFG 1.0 ignores negative prompt. Don't use it if you need negative)

The best coherency: use CFG 2.8 and CFG_slider lora on +0.75 weight (for Euler A).