<h1 id="haruka-sakura-or-lora-wind-breaker-or-base:-anima">Haruka Sakura | LoRA — WIND BREAKER | Base: ANIMA</h1>🙏 Before anything else: this LoRA took a lot of work — picking the images, building the dataset and tags, training, and testing the checkpoints. If you have a moment, drop a Like and especially post your results in the comments (with the prompt/settings you used) — it helps me a lot to know what to improve in the next versions. 🔥⚠️ Contains limited NSFW capability (see details below). Use it respectfully. 💬 Check out how the prompt examples were set up in the posted images. 💬 Leave your generations in the comments, and tag the LoRA — it would help a lot to see how it's looking. 📦 Trained with 172 diverse images at 1024px. Releasing two checkpoints (600 and 800 steps).<hr /><h2 id="about-this-model">📊 About this model</h2>This LoRA captures Haruka Sakura's very distinct design, which is usually hard to get right in the standard ANIMA base. It accurately reproduces her split hair (black on one side, white on the other) and her heterochromia (one golden-yellow eye, the other purple-gray). It also does a great job nailing the Furin High School uniform (Bofurin jacket) and her characteristic facial expressions, ranging from fierce battle faces to that easily embarrassed/clumsy look.Compared to the base model, both checkpoints significantly improve character likeness and bring a much cleaner anime aesthetic.⚠️ Dataset &amp; NSFW Warning: the training dataset consists of 172 images (85% anime screenshots and 15% anatomy-focused fan art containing NSFW). The model reproduces the character extremely well when clothed. However, because the NSFW training subset is small, generating NSFW content might result in anatomical or proportional errors if your prompt isn't carefully structured. Keep this in mind when prompting.<hr /><h2 id="generation-tips">ℹ️ Generation tips</h2><ul><li>On ANIMA base with 12 steps + acceleration LoRA, the output keeps a clean anime feel that works well for Sakura — recommended as a starting point</li><li>Using more steps or a different ANIMA-based model with quality LoRAs will improve sharpness and detail</li><li>Higher CFG can help lock in her dual hair color and heterochromia more consistently</li><li>Step 600: the most versatile checkpoint. It responds better to prompts, makes it easier to change her outfits/backgrounds, and is simpler to use overall</li><li>Step 800: provides higher likeness and cleaner line art, but becomes a bit more rigid with outfit swaps or complex prompts. Use if you want maximum accuracy in her standard school uniform</li><li>Check the prompt examples used in the posted preview images — they show how to structure tags for her look</li></ul><hr /><h2 id="checkpoints">📊 Checkpoints</h2>StepsResult⭐ 600 Recommended: best balance of style and flexibility. Easier to use, better prompt adherence, and very versatile for different outfits.800 High Quality: better character likeness and sharper anime style. Slightly more rigid, but ideal if you want the highest quality in her standard look.Both checkpoints are available, so you can pick the one that best fits your specific prompt and styling needs.<hr /><h2 id="generation-settings-used-for-preview-images">⚙️ Generation settings used for preview images</h2>All preview images were generated using:<ul><li>Base model: Anima 1.0</li><li>Acceleration LoRA: strength 1.0</li><li>CLIP: anima default</li><li>Steps: 12</li><li>CFG: 1.0</li><li>Sampler: <code>er_sde - beta</code></li></ul><hr /><h2 id="one-more-thing">👍 One more thing</h2>I'm just getting started with this — if you have a moment, drop a Like and share your generations with the settings you used. It really helps me improve the next versions. Have fun! 🔥

Files with SHA256: 7a5382e79cff85530ebd6da1c7a1b1f20adb131aab67471e8a2f68c40cbe8bde