LineartAnime Redmond ZImageTurbo LoRA

• Clean anime lineart — Trained specifically for consistent, professional lineart style illustrations
• Trigger words: "Lineart, LineAniAF" — Add these to your prompt to activate the style
• Base model: Z Image Turbo — Lightweight adapter (81 MB) for Tongyi-MAI/Z-Image-Turbo

What This Model Does

LineartAnime Redmond ZImageTurbo is a LoRA (Low-Rank Adaptation) that teaches the Z Image Turbo model to generate clean, manga-style lineart illustrations. Unlike full checkpoints, this 81 MB adapter layers on top of the base Z Image Turbo model, adding the ability to produce crisp black-and-white line drawings with anime aesthetics.

The model excels at character illustrations — young women, schoolgirls, warriors, and traditional Japanese characters — all rendered with confident linework and soft shading typical of manga art.

Training Background

This LoRA was trained using GPU compute sponsored by Redmond.AI. The base model is Tongyi-MAI/Z-Image-Turbo, a fast text-to-image model. Training focused specifically on lineart style transfer, resulting in an adapter that consistently produces clean lines without the artifacts common in some anime-style LoRAs.

License: Apache-2.0 — free for commercial and non-commercial use.

Trigger Words and Usage

Include these trigger words in your prompt:

Lineart, LineAniAF

Example prompts from the model card:

• Lineart, LineAniAF young woman playing acoustic guitar, park bench, autumn leaves falling, peaceful moment, soft shading, manga style
• Lineart, LineAniAF schoolgirl with long hair, classroom background, books on desk, sunlight through window, nostalgic vibe, clean lines
• Lineart, LineAniAF warrior girl with sword, standing pose, dynamic angle, fantasy armor, flowing cape, action shot, bold linework
• Lineart, LineAniAF girl in traditional kimono, cherry blossom garden, petals floating, serene expression, Japanese style, detailed patterns

Recommended: Use with ComfyUI for optimal results. The LoRA works with standard Stable Diffusion pipelines that support Z Image Turbo.

Recommended Settings

Parameter | Recommendation

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

Base Model | Tongyi-MAI/Z-Image-Turbo LoRA Weight | 0.8–1.0 (start at 1.0) Sampler | Euler a / DPM++ 2M Karras Steps | 20–30 CFG Scale | 5–7 Resolution | 1024×1024 or 512×768 Adjust LoRA weight downward if lines become too heavy or style bleeds into unwanted areas.

Common Errors and Fixes

Error | Cause | Fix | Source

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

Style not activating | Missing trigger words | Add "Lineart, LineAniAF" to prompt | HF Model Card Lines too thick/heavy | LoRA weight too high | Reduce weight to 0.7–0.8 | Community testing Color bleeding into lineart | Base model influence | Increase CFG, use negative prompt "color, grayscale, shading" | HF Model Card Inconsistent line quality | Low steps / wrong sampler | Use 25+ steps, Euler a or DPM++ 2M Karras | HF Model Card

FAQ

Q: Does this work with other base models like SDXL or Flux? A: No. This LoRA is trained specifically for Z Image Turbo (Tongyi-MAI/Z-Image-Turbo). Using it with other architectures will produce broken results.

Q: Can I use this commercially? A: Yes, the Apache-2.0 license permits commercial use, modification, and distribution.

Q: What's the difference between this and the FLUX version? A: This is for Z Image Turbo only. ArtificialGuyBR publishes separate LineartAnime LoRAs for FLUX, Qwen-Image, and SDXL — each trained on its respective base model.

Q: Why are my images coming out with color instead of pure lineart? A: Z Image Turbo is a full-color model. Add negative prompts like "color, grayscale, shading, watercolor" and ensure trigger words are at the start of your prompt.

Sources

• HuggingFace Model Card — Official model card with trigger words, example prompts, and license
• Civitai Model Page — Community stats, alternative versions (FLUX, Qwen), and user reviews
• Redmond.AI — GPU compute sponsor for training
• Z Image Turbo Base Model — Base model this LoRA adapts