<h3 id="test-23">Test 23</h3><h3 id="trigger-words:-m1er0sk">Trigger words: m1er0sk</h3>And"dense cross hatching". However, prompts such as "hatching," "sketch," or "traditional media" significantly affect the image, so use them with caution.<hr /><h3 id="my-pencil-sketch-style-lora-(2025-version-and-2026-consolidated-version)">My Pencil Sketch Style LoRA (2025 Version &amp; 2026 Consolidated Version)</h3>Future updates for the Anima version of my sketch LoRA will be posted on this page. <ul><li>2025 Version: It seems the 2025 version generally outperforms the 2026 version.</li><li>2026 Version (Consolidated): It performs slightly better in handling exaggerated shapes and rendering realistic prompts.</li></ul><h3 id="the-training-results-are-truly-hard-to-control.">The training results are truly hard to control.</h3> I’ve posted comparison images (including the May version) in the image section. I’m going to look into how to train the Krea 2 model now; hopefully, it won't be such a headache.<hr /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f7cde7af-f0d7-4681-986a-5605c3ee5ae6/original=true/f7cde7af-f0d7-4681-986a-5605c3ee5ae6.jpeg" />（This image ↑ shows some exercises that borrow composition from "Shunga". These are not in the training dataset. Perhaps once I have produced enough of these types of paintings, I will train a dedicated model.）

Files with SHA256: 6fb7c971ae29acdaa9651901469afecae5371a25e1d018436eb612c0a2bfe556