If you like what I do here, please consider supporting my work via Ko-fi &lt;3 <a target="_blank" rel="ugc" href="https://ko-fi.com/pixelworldai">https://ko-fi.com/pixelworldai</a>Introducing PiXL, an SDXL model with a pronounced anime/cartoon bias!I aimed to finetune a model so that minimal prompting would be required to achieve an anime/cartoon/western comic style. With PiXL, you can achieve just that!After testing 100 random prompts, focusing solely on subject matter without any mention of style, I found that 88/100 of them had an anime/western comic style.Naturally, the results improve when specific style prompts are used. However, with PiXL, achieving these styles requires very few prompts (or more for added style!)Here are some sample style prompts you might want to mess with:simple anime style, 2D, celshading, thick lineart, heavy black ink lines, posterized, flat color, toonshading, Cartoon Rendering, Flat Shading, Graphic Novel Style, minimal shading, hard color shading.This model was finetuned with a 100% AI generated dataset, with <a target="_blank" rel="ugc" href="https://civarchive.com/user/braintacles">braintacles</a> wonderful <a target="_blank" rel="ugc" href="https://civarchive.com/models/131703/brixl-or-a-must-in-your-toolbox">BriXL</a> model as it's base. Huge thanks to him for helping me get this together! &lt;3The first 10 images in my example are done with A1111, settings are in the metadata and are pretty standard. The last 10 images were made in ComfyUI using the Turbo Scheduler, and since I can't embed the metadata here are the settings:Steps: 10CFG: 4Sampler: dpmpp_3m_sdeUsing the turbo scheduler is my favorite way of using this model, the results are clean (enough) and it is really fast! If you run an upscale workflow in Comfy, I highly recommend your first generation step be with turbo, then upscale with your usual ksampler settings.This model still produces realistic images, but as I said earlier, it leans more toward toon-type images. Future versions of this model will hopefully have more of a set anime/cartoon style with less prompting to get there.Hope you enjoy! :)

PiXL - Anime/Cartoon/Comic Style

pixlAnimeCartoonComic_v10.safetensors

v1.0

If you like what I do here, please consider supporting my work via Ko-fi &lt;3 <a target="_blank" rel="ugc" href="https://ko-fi.com/pixelworldai">https://ko-fi.com/pixelworldai</a>Introducing PiXL, an SDXL model with a pronounced anime/cartoon bias!I aimed to finetune a model so that minimal prompting would be required to achieve an anime/cartoon/western comic style. With PiXL, you can achieve just that!After testing 100 random prompts, focusing solely on subject matter without any mention of style, I found that 88/100 of them had an anime/western comic style.Naturally, the results improve when specific style prompts are used. However, with PiXL, achieving these styles requires very few prompts (or more for added style!)Here are some sample style prompts you might want to mess with:simple anime style, 2D, celshading, thick lineart, heavy black ink lines, posterized, flat color, toonshading, Cartoon Rendering, Flat Shading, Graphic Novel Style, minimal shading, hard color shading.This model was finetuned with a 100% AI generated dataset, with <a target="_blank" rel="ugc" href="https://civitai.com/user/braintacles">braintacles</a> wonderful <a target="_blank" rel="ugc" href="https://civitai.com/models/131703/brixl-or-a-must-in-your-toolbox">BriXL</a> model as it's base. Huge thanks to him for helping me get this together! &lt;3The first 10 images in my example are done with A1111, settings are in the metadata and are pretty standard. The last 10 images were made in ComfyUI using the Turbo Scheduler, and since I can't embed the metadata here are the settings:Steps: 10CFG: 4Sampler: dpmpp_3m_sdeUsing the turbo scheduler is my favorite way of using this model, the results are clean (enough) and it is really fast! If you run an upscale workflow in Comfy, I highly recommend your first generation step be with turbo, then upscale with your usual ksampler settings.This model still produces realistic images, but as I said earlier, it leans more toward toon-type images. Future versions of this model will hopefully have more of a set anime/cartoon style with less prompting to get there.Hope you enjoy! :)

Files with SHA256: c6b91b7c3b7743e1d7635bf2f30c22597d137dc86f519eda80ba551cf9428550