ghost
Sign In

Aries Spring (Astra Lost in Space / Kanata no Astra)

57
332
51
4
Verified:
SafeTensor
Type
LoRA
Stats
332
51
Reviews
Published
Aug 8, 2023
Base Model
SD 1.5
Training
Steps: 1,380
Epochs: 8
Usage Tips
Clip Skip: 2
Trigger Words
aries_spring
pink hair, short hair, (hair ornament, heterochromia:1.5), green eyes, yellow eyes
white bodysuit, pink gloves
galaxy, spacecraft, starry sky
Hash
AutoV2
381AD5B1D0
default creator card background decoration
allureAI's Avatar
allureAI

Aries Spring (Astra Lost in Space / Kanata no Astra)

Please feel free to rate the model or share your creations with it. I'd be thrilled to see your works. If you find it useful, consider showing your appreciation by buying me a coffee.

  • Weight: between 0.8 to 1.0

  • Trigger Word: aries_spring

Additional Prompts:

  • Character: aries_spring, pink hair, short hair, hair ornament, heterochromia, green eyes, yellow eyes

  • Spacesuit: white bodysuit, pink gloves

  • Space Background: galaxy, spacecraft, starry sky

Cutoff Settings (Recommended):

  • Target tokens (character only): yellow,green,pink

  • Target tokens (character with white bodysuit): yellow,green,pink,white,pink

  • Weight: 0.5

Suggested Hires.fix configuration:

  • Upscaler (For Anime Models): R-ESRGAN 4x+ Anime6B

  • Upscaler (For PhotoReal Models): R-ESRGAN 4x+

  • Hires steps: between 10 to 20

  • Denoising strength: between 0.15 to 0.35

This character is characterized by heterochromia, possessing one yellow eye and one green eye. I discovered that using a cutoff extension helps render them more consistently across various base models. An alternative approach to achieve stability is by increasing the weight of the prompts specifically related to the yellow eyes, green eyes, and heterochromia. This adjustment ensures a more accurate and balanced representation of the character's unique eye colors.

Your comments or suggestions are not only welcome but can significantly improve the quality of my models. They also motivate me to create more LoRA models. Feel free to connect with me on my Discord channel.

V1 was trained on the AnyLoRA base model with a dataset comprised of 33 images.

The compatibility of this LoRA model has been tested and work well with the following base models:

Anime models:

  • MeinaMix V11

  • DreamShaper V8

  • Perfect World V5

Photorealistic models:

  • AbsoluteReality V1

  • MoonMix V3 Reality

Cutoff Installation

Automatic1111 SD WebUI -> Extensions -> Install from URL

URL for extionsion's git repo: https://github.com/hnmr293/sd-webui-cutoff

Or, from Available tab search "cutoff".

Then click "Install" button.

For ComfyUI, please follow this guide.

All demonstration images showcased are randomly selected, not handpicked for optimal results.