ArtificialGuyBR

Home / Blog / Training LoRAs

LineAniRedmond LinearMangaSDXL V2: Manga Lineart LoRA for SDXL

Generate manga-style lineart with this SDXL LoRA. Trigger words, style tags, recommended settings, and example images.

2 sources cited Training LoRAs

On this page

  1. What This Model Does
  2. Trigger Words & Style Control
  3. Recommended Settings
  4. How to Use
  5. License
  6. Common Errors and Fixes
  7. FAQ
  8. Sources
  9. Related Models

LineAniRedmond LinearMangaSDXL V2: Manga Lineart LoRA for SDXL

TL;DR

What This Model Does

LineAniRedmond LinearMangaSDXL V2 is a Low-Rank Adaptation (LoRA) trained on top of Stable Diffusion XL 1.0 (stabilityai/stable-diffusion-xl-base-1.0) to produce manga-style lineart. It was developed by ArtificialGuyBR with GPU compute sponsored by Redmond.AI and marks the author's first anime-focused LoRA release.

The LoRA excels at generating character illustrations with the clean, expressive linework characteristic of Japanese manga — from detailed cross-hatching to minimalist single-line styles.

Manga lineart character example showing clean linework and anime proportions

Trigger Words & Style Control

Tag | Effect

|-----|--------

LineAniAF, lineart | Required trigger — activates the LoRA style detailed | Adds cross-hatching, shading lines, intricate details minimalist | Reduces to essential contour lines only colorful | Encourages flat color fills within lineart black and white | Forces monochrome output Combine tags for hybrid styles (e.g., LineAniAF, lineart, detailed, black and white).

The author notes the LoRA "sometimes needs more than one gen to create good images" — expect to reroll a few times for complex prompts.

How to Use

ComfyUI

Automatic1111 / WebUI

  1. Place LineAniRedmondV2-Lineart-LineAniAF.safetensors in models/Lora
  2. Add <lora:LineAniRedmondV2-Lineart-LineAniAF:0.75> to prompt
  3. Include trigger words: LineAniAF, lineart

Diffusers (Python)

from diffusers import StableDiffusionXLPipeline
import torch

pipe = StableDiffusionXLPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    torch_dtype=torch.float16,
    variant="fp16"
).to("cuda")

pipe.load_lora_weights(
    "artificialguybr/LineAniRedmond-LinearMangaSDXL-V2",
    weight_name="LineAniRedmondV2-Lineart-LineAniAF.safetensors"
)

prompt = "LineAniAF, lineart, detailed, anime girl, cherry blossoms"
image = pipe(prompt, num_inference_steps=25, guidance_scale=6).images[0]

License

CreativeML OpenRAIL-M — permits commercial and non-commercial use. You must include the license and attribution when distributing the model or derivatives. See the license text for full terms.

Common Errors and Fixes

Error | Cause | Fix

|-------|-------|-----

Output looks like base SDXL, no lineart | LoRA not loaded or weight too low | Verify LoRA filename, increase weight to 0.75–0.9 Over-saturated lines, artifacts | LoRA weight too high | Reduce to 0.5–0.6 Prompt ignored, generic anime face | Missing trigger words | Always include LineAniAF, lineart Black-and-white tag produces color | Tag placement | Move black and white earlier in prompt

FAQ

Q: Is this compatible with SD 1.5? A: No. This LoRA is trained specifically for SDXL 1.0. It will not work with SD 1.5 models.

Q: Can I merge this LoRA into a checkpoint? A: Yes, using tools like sd-scripts or the A1111 "Checkpoint Merger" tab. Merge at 0.7–0.8 alpha for best results.

Q: Does it work with Pony Diffusion or other SDXL finetunes? A: It can work but results vary. The LoRA was trained on base SDXL 1.0; finetunes may shift the style unpredictably.

Q: Where can I test it online? A: The author provides a demo space: artificialguybr-demo-lora

Sources