<h1 id="mellow_pencil-xl_clear">mellow_pencil-XL_clear</h1><ul><li>Guide v2 (External site): <a target="_blank" rel="ugc" href="https://www.ai-image-journey.com/2026/04/mellow-pencil-xl-clear-v2.html">English</a> | <a target="_blank" rel="ugc" href="https://note.com/ai_image_journey/n/n834efb241a1a">Japanese</a></li><li>Guide v1 (External site): <a target="_blank" rel="ugc" href="https://www.ai-image-journey.com/2025/11/mellow-pencil-xl-clear.html">English</a> | <a target="_blank" rel="ugc" href="https://note.com/ai_image_journey/n/nb87c8cea4b05">Japanese</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/easygoing0114/mellow_pencil-XL_clear">Hugging Face</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2171491/sdxlanimeclearvae">SDXL_anime_clear_vae</a></li></ul><h2 id="mellow_pencil-xl_clear_v2">mellow_pencil-XL_clear_v2</h2>mellow_pencil-XL_clear_v2 | mellow_pencil-XL_clear_v1<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/e87cfa8a-4783-4102-a29d-6f6d519f7a44/original=true/e87cfa8a-4783-4102-a29d-6f6d519f7a44.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ff825a0c-d513-4ce6-a327-b297d873e19e/original=true/ff825a0c-d513-4ce6-a327-b297d873e19e.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/29d45f68-3711-4df5-b925-da8917bdd93e/original=true/29d45f68-3711-4df5-b925-da8917bdd93e.jpeg" />The previous mellow_pencil-XL_clear_v1 was a model specifically enhanced with intermediate noise to create a sharp, high-definition output.While this increased contrast and detail, it sometimes resulted in a "burnt" color look or left behind color blotches/unevenness caused by noise artifacts.In mellow_pencil-XL_clear_v2, I have slightly reduced the intermediate noise to achieve natural color expression and a cleaner texture. As a trade-off, the density of fine background details is slightly lower compared to v1.<hr /><h2 id="int8_convrot_hq-(light-weight)">INT8_ConvRot_HQ (light weight)</h2>INT8_ConvRot_HQ is a lightweight version. The CLIP and VAE remain unchanged, while only the intermediate layers of the UNET have been converted to INT8_ConvRot format.Image quality is slightly lower than the original v2, but it is particularly well-suited for fast generation in low-VRAM environments (4-6 GB).INT8_ConvRot is currently supported in ComfyUI and Stable Diffusion webUI Forge NEO.<hr /><h3 id="recommended-settings-(v2)">Recommended Settings (v2)</h3><ul><li>CFG scale: 3 (2-4)</li><li>Sampler: euler a</li><li>Scheduler: beta</li><li>Steps: 12-20 (around 15)</li></ul>Quality Tags and Negative Prompts are not always necessary.<h3 id="merge-recipe-(v2)">Merge Recipe (v2)</h3><ul><li>UNET: <a target="_blank" rel="ugc" href="https://huggingface.co/bluepen5805/mellow_pencil-XL/tree/main">mellow_pencil-XL-v1.0.0-base</a><ul><li>input_blocks.0: x0.85</li><li>output_blocks.6: x1.1</li><li>out.: x1.05</li></ul></li><li>CLIP: <a target="_blank" rel="ugc" href="https://huggingface.co/bluepen5805/mellow_pencil-XL/tree/main">mellow_pencil-XL-v1.0.0-base</a></li><li>VAE: <a target="_blank" rel="ugc" href="https://huggingface.co/easygoing0114/SDXL_anime_clear_vae">SDXL_anime_clear_vae</a></li></ul><hr /><h2 id="mellow_pencil-xl_clear_v1">mellow_pencil-XL_clear_v1</h2>mellow_pencil-XL-v1.0.0-base_clear | mellow_pencil-XL-v1.0.0-base (Original)<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1c0b8be4-f27b-41fa-a708-6ad778260501/width=525/1c0b8be4-f27b-41fa-a708-6ad778260501.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a931d8d4-7003-4c4c-aaf2-5bb0ec6f10b8/width=525/a931d8d4-7003-4c4c-aaf2-5bb0ec6f10b8.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/368c157e-0f33-4286-9ab1-0a51fc19ce15/width=525/368c157e-0f33-4286-9ab1-0a51fc19ce15.jpeg" />Fine Tuned Model of <a target="_blank" rel="ugc" href="https://huggingface.co/bluepen5805/mellow_pencil-XL">mellow_pencil-XL</a> series, reduce noise for a clearer look and ensures vivid color expression even when the CFG scale is set low.<h3 id="recommended-settings-(v1)">Recommended Settings (v1)</h3><ul><li>CFG scale: 3 (3-5)</li><li>Sampler: euler a</li><li>scheduler: beta</li><li>Steps: 15-25 (around 20)</li></ul>Quality Tags and Negative Prompts are not always necessary.<h3 id="merge-recipe-(v1)">Merge Recipe (v1)</h3><ul><li>UNET: <a target="_blank" rel="ugc" href="https://huggingface.co/bluepen5805/mellow_pencil-XL/tree/main">mellow_pencil-XL-v1.0.0-base</a><ul><li>time_embed.: x0.90</li><li>input_blocks.0: x0.90</li></ul></li><li>CLIP: <a target="_blank" rel="ugc" href="https://huggingface.co/bluepen5805/mellow_pencil-XL/tree/main">mellow_pencil-XL-v1.0.0-base</a><ul><li>clip_l.encorder.layers.4: x0.95</li><li>clip_g.encorder.layers.13: x0.95</li><li>clip_g.encorder.layers.31: x0</li></ul></li><li>VAE: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2171491/sdxlanimeclearvae">SDXL_anime_clear_vae</a></li></ul><hr /><h2 id="upscaler">Upscaler</h2>your choice, examples:<ul><li><a target="_blank" rel="ugc" href="https://openmodeldb.info/models/4x-realesrgan-x4plus-anime-6b">RealESRGAN_x4Plus Anime 6B</a> (Commercial use)</li><li><a target="_blank" rel="ugc" href="https://openmodeldb.info/models/4x-IllustrationJaNai-V1-ESRGAN">IllustrationJaNai_V1_ESRGAN</a> (No-Commercial use)</li></ul><hr /><h2 id="detailer">Detailer</h2>works well.<hr /><h2 id="tips">Tips</h2><ul><li>Color Prompts: <a target="_blank" rel="ugc" href="https://www.ai-image-journey.com/2025/10/dyes-color-codes.html">English</a> | <a target="_blank" rel="ugc" href="https://note.com/ai_image_journey/n/nbe7eeb3632c3">Japanese</a></li><li>Color Correct custum nodes in ComfyUI: <a target="_blank" rel="ugc" href="https://www.ai-image-journey.com/2025/09/comfyui-easygoing-nodes.html">English</a> | <a target="_blank" rel="ugc" href="https://note.com/ai_image_journey/n/n5bb33311b866">Japanese</a></li><li>About VAE: <a target="_blank" rel="ugc" href="https://www.ai-image-journey.com/2026/04/sdxl-anime-natural-vae.html">English</a> | <a target="_blank" rel="ugc" href="https://note.com/ai_image_journey/n/nd22f8c4cabc7">Japanese</a></li><li>Illustrious-XL series License: <a target="_blank" rel="ugc" href="https://www.ai-image-journey.com/2025/10/sdxl-license.html">English</a> | <a target="_blank" rel="ugc" href="https://note.com/ai_image_journey/n/n4982b466f663">Japanese</a></li></ul><hr /><h2 id="license">License</h2>The mellow_pencil-XL_clear model released here is subject to the restrictions of both of the following two licenses:<ul><li><a target="_blank" rel="ugc" href="https://freedevproject.org/faipl-1.0-sd/">Fair AI Public License 1.0-SD</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/spaces/CompVis/stable-diffusion-license">CreativeML Open RAIL-M</a></li></ul>The key points of these two licenses are the requirement to publish the recipe when releasing a derivative model, and furthermore, that Onoma AI, the source of Illustrious-XL 2.0, retains the right to modify the derivative model and its outputs.We have <a target="_blank" rel="ugc" href="https://huggingface.co/OnomaAIResearch/Illustrious-XL-v2.0/discussions/4">inquired with Onoma AI</a> regarding the detailed license for derivative models of Illustrious-XL 2.0, but until a response is received, both of the above restrictions shall apply.

Files with SHA256: 3e948c0b9dc1bc90d54377366d9bc72e1cb1a4c216263008501a88d2e25bd1c5