Sign In

Mitya (Genshin Impact) | Anima

0

Updated: Nov 15, 2026

character

Not recently loaded

Download

1 variant available

bf16 SafeTensor

mitya_anima_v1.safetensors

BF16, good balance • 43.88 MB

Verified:

Type
LoRA
Stats

8

0

Reviews
Published

Nov 15, 2026

Base Model
Training
Steps: 1,840Epochs: 20
Usage Tips
Strength: 1
Hash
AutoV2602FA19024
Trigger Words
mitya \(genshin impact\)

The CircleStone Model is licensed by CircleStone Labs LLC under the CircleStone Non-Commercial License. Copyright CircleStone Labs LLC. IN NO EVENT SHALL CIRCLESTONE LABS LLC BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.

Built on NVIDIA Cosmos

pv14.png

Character LoRA for Mitya from Genshin Impact, trained on Anima base 1.0.

The outfit is baked into the trigger: light blue hair with an ahoge, violet eyes, high violet collar, white cravat with an atom brooch, golden rings on the shoulders, and the wide teal-lined cape with tassels. You do not need to describe his appearance in the prompt.

Usage

- Trigger: mitya \(genshin impact\) (keep the backslashes before the brackets)

- Recommended weight: 1.0 (all previews were made at 1.0)

- Sampler: euler, simple scheduler, 30 steps, CFG 4

- Do not use together with the Anima turbo/distillation LoRA at low step counts without testing, as I only tested with the base model.

Prompting

Anima works best with tags plus a short natural-language sentence, on one line:

mitya \(genshin impact\), genshin impact, 1boy, solo, upper body, looking at viewer, gentle smile. Mitya reading a book in an old library, warm light through tall windows.

Artist tags with @ work and the style is preserved (for example @mikami saka). The previews use style mixes of several artist tags with watercolor and faux traditional media tags.

Alternate outfits

Describe the clothes in the prompt and the LoRA follows them. In my test, "white hoodie" produced a plain hoodie with no collar, brooch or rings leaking in.

Known limits

- Trained mostly on official art and fan art, so game-style 3D renders are not a strength.

- Full-body and back views are less consistent than portraits and upper body.

- Details of the cape tassels vary between generations.

- I did not train or test any scarf or accessories not in his default outfit.

Training

- 92 images, kohya sd-scripts, Anima base 1.0

- Network dim 16 / alpha 8, LR 5e-5, AdamW8bit, cosine, 20 epochs, batch 2, 1024 buckets

- Text encoder not trained (unet only)

License: Anima is a non-commercial model. Generated images follow the base model's license terms.