<h2 id="gym-onee-san-from-otakuvs"><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fa7d627c-1c0e-4bab-b38d-e314e4d77131/original=true/fa7d627c-1c0e-4bab-b38d-e314e4d77131.jpeg" />✨ Gym Onee-san from OtakuVS</h2>— A sporty girl who appears in a short skit with Peepo-chan. Peepo casually offers her a ride in her car, only for the whole thing to fall apart when she realizes she has no idea where she parked it. She only gets a brief appearance, but her relaxed, slightly confused energy makes the scene pretty memorable.<h2 id="greaterotakuvs-collectionless"><a target="_blank" rel="ugc" href="https://civarchive.com/collections/10941612">&gt;OtakuVS COLLECTION&lt;</a></h2><h1 id="turn-off-your-rxxxx-filter-to-see-hidden-juicy-loras-in-greaterhereless">⚠️Turn off your R/X/XXX Filter to see hidden JUICY Loras in <a target="_blank" rel="ugc" href="https://civarchive.com/collections/13797500">&gt;HERE&lt;</a></h1><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/e4c11a1c-1328-4124-9572-eee4f03663f0/original=true/e4c11a1c-1328-4124-9572-eee4f03663f0.jpeg" /><h3 id="how-i-generate-my-images">🖼️ How I Generate My Images</h3>I made a full article that explains how I generate my images — including my settings, prompting tips, and even my potato PC setup 😅👉 <a target="_blank" rel="ugc" href="https://civarchive.com/articles/17562">Check it out here</a> if you’re curious or want to try the same method.<h3 id="nsfw-gallery-(18+)">🔞 NSFW gallery (18+)</h3>Check my account <a target="_blank" rel="ugc" href="https://sociabuzz.com/zetsubou">here </a>to see the spicy showcase that I can't show here, if any.<h3 id="prompt-tips">🎨 Prompt Tips</h3>These are optional starting points for handling a few behaviors of this version. You may still need some retries depending on the checkpoint, prompt, and LoRA strength.<h3 id="lora-strength-and-design-flexibility">⚖️ LoRA strength and design flexibility</h3>This LoRA is still influenced by its limited source coverage, so features that are not clearly defined in the prompt may fall back toward the learned default interpretation.A LoRA weight of <code>1</code> can still work well, especially when the design details you want are clearly described.Only lower the LoRA weight gradually if default details keep leaking even after using clear positive and negative tags. Lowering it too much can also weaken the character’s face, hair, and overall identity.<h2 id="lora-project">📊 Lora Project</h2>If you want to know what characters that I'm covering next you can check these link : 👉 <a target="_blank" rel="ugc" href="https://docs.google.com/spreadsheets/d/1oM6Xk0k5sBCw2CQfIFMzQQuuC6lHdOOnw3zES9phBm8/edit?gid=1864372839#gid=1864372839">Personal Project</a> 👉 <a target="_blank" rel="ugc" href="https://docs.google.com/spreadsheets/d/1k15ya9i0TjAHJnBpVfPSsjHsngpjkLsAZ11IcOT0hc0/edit?gid=0#gid=0">Commision Project</a>If this model sparked something in you, show some love with a like, review, or buzz. Now go— Be wild

Gym Onee-san | OtakuVS

🎬 Trained from a very small source set with AI-assisted extension<pre><code>This LoRA was trained from 4 usable reference images taken from the YouTube video series, which were then extended with AI assistance to make the dataset more practical for training.

Because the original source coverage is still very limited, this version is built around a small set of references rather than a wide range of scenes, poses, or outfit situations.</code></pre>🧥 Designed for flexible hair and jacket behavior<pre><code>This version was intentionally designed to keep her hair and jacket from becoming too heavily baked into the character.

The goal was to make those features more flexible during generation, including allowing the jacket to be used more interactively, such as wearing it open or closed instead of treating it as one fixed state.</code></pre>🎨 Overcast source images were color corrected<pre><code>In the original reference images, she appears under an overcast sky, which makes her skin tone look a bit darker due to the cloudy lighting conditions.

To reduce that lighting bias, I color corrected the dataset so the LoRA would not learn the darker overcast tint as if it were her actual skin tone.</code></pre>As usual, the final result can still depend on your checkpoint, prompt, LoRA strength, CFG scale, camera angle, and generation settings.

Files with SHA256: ac5e4deb1e6e13f4d4b1a7d79dfda6bfa73735207ce19ac4dd3205eea0790e94