ASTX-Anima_2.9B
🎨 About this Model
A much-suffered and long-tested model, but in the end it turned out very solid: good detail, prompt comprehension, attention to small things.
Essentially a regular Anime model, but with its own stylization, with several LoRAs I trained baked in.
📦 3 versions will be released
⚡ As well as 2 quants
UPD 05/11/26:
v3.0 i will add to 2.0 Release, because not so much diff in this versions.
Only small changes with drawing vision, so no need to set it like new release.
⚙️ Settings
🎬 My settings for Anima
📸 CLIP Skip: 2
🎲 CFG Scale: 3 – 3.5 (max)
🧼 Sampler / Scheduler: Res_Multistep or ER_SDE with SGM_Uniform or Beta, or use Simple ⭐ best way!
👣 Steps: 30 – 35 (max)
♨️ Hi-Res Fix: ❌ I am not using it
🎭 ADetailer: ❌ I am not using it
📐 Recommended Resolution
1024×1536 / 1536×1024
832×1216 / 1216×832
Simple: 1024×1024
✨ Prompt Parameters
✔️ Quality (Anima): not needed.
❌ Negative (Illustrious + Anima):
score_1, score_2, score_3, blurry, jpeg artifacts, lowres, worst quality, low quality, bad hands, bad anatomy, signature, watermark, patreon logo, patreon username, twitter username, copyright name, twitter logo, artist name, simple backgroundDescription
ASTX-Anima_2.9B — V2.0
🎨 About this Model
In this version, the influence of LoRA is significantly reduced. LoRA for styles also has a weak influence on this model.
The goal of this model is simple — maximum quality, detail, and precision. Overall there's room to improve, but I like what came out.
I didn't use a single LoRA in my preview images, no modifications or other crap: just the prompt, standard resolution, no upscaling, and no ADetailer.
📦 3 versions will be released
⚡ As well as 2 quants
⚙️ Settings:
🎬 My settings for Anima
📸 CLIP Skip: 2
🎲 CFG Scale: 3 – 3.5 (max)
🧼 Sampler / Scheduler: custom preset: Res_Multistep or ER_SDE with SGM_Uniform or Beta, or use Simple ⭐ best way!
👣 Steps: 30 – 35 (max)
♨️ Hi-Res Fix: ❌ I am not using it
🎭 ADetailer: ❌ I am not using it
📐 Recommended Resolution
1024×1536 / 1536×1024
832×1216 / 1216×832
Simple: 1024×1024
✨ Prompt Parameters
✔️ Quality (Anima): not needed.
❌ Negative (Illustrious + Anima):
score_1, score_2, score_3, blurry, jpeg artifacts, lowres, worst quality, low quality, bad hands, bad anatomy, signature, watermark, patreon logo, patreon username, twitter username, copyright name, twitter logo, artist name, simple background