<h1 id="v2-notes:-jbkmq3x1d">V2 Notes:</h1><ul><li>I poison it with my AGM Lora</li></ul>Read Me Please on How to Use 🦄 <code>MistlanMix</code> is a versatile Pony Model that has a natural <code>2D</code> anime look but can also do many other styles without having to utilize style LoRAs. It mainly focuses on creating beautiful detailed backgrounds and poses all in one.Using descriptions from sample images above to assist such as: <code>bright colors</code>, <code>dark background</code>, <code>vivid lighting</code><h3 id="how-to-use-correctly:-h0w0kavde">How to Use correctly:</h3>Use: `score_9`, `score_8_up`, `score_7_up`, `rating_explicit`, (if newds)Use Clip Skip 2. - Using the `danbooru` tags when prompting.<ul><li>Such as <code>anime_screencap</code> for 2d anime look in positive</li><li>Or <code>realism, 3d </code>for a realistic 2.5 look</li></ul>Additional keywords improve output such as: `realistic`, `masterpiece`, `scenery`, `beautiful eyes`, `best quality`, `very aesthetic`.- When prompting, Using less is better for output!!!- Recommended aspect ratios are `1024x1024`, `832x1216`, `1216x832`. Works flawlessly with many SDXL resolutions.- Highly Recommended to use SDFORGE (if lower VRAM) using a hires fix to txt2img outputs<ul><li>Hires.fix Upscaler: R-ESRGAN 4x+ Anime6B</li><li>Upscale 1.2-2</li><li>Hires Steps 20-25</li><li>Sampling Steps 25-40</li></ul><h3 id="difficulties:-z4zb9ktxm">Difficulties:</h3><ul><li>Sometimes generating multiple fingers and toes</li></ul><h3 id="mixes:-x0sfrr45q">Mixes:</h3>TBA<h3 id="improvements-goals-qpvaxflhh">Improvements Goals</h3><ul><li>better background</li><li>better prompting</li><li>less wording</li></ul>

Pony MistlanMix PDXL

ponyMistlanmixPDXL_v10.safetensors

v1.0

Read Me Please on How to Use 🦄 <code>MistlanMix</code> is a versatile Pony Model that has a natural <code>2D</code> anime look but can also do many other styles without having to utilize style LoRAs. It mainly focuses on creating beautiful detailed backgrounds and poses all in one.Using descriptions from sample images above to assist such as: <code>bright colors</code>, <code>dark background</code>, <code>vivid lighting</code><h3 id="how-to-use-correctly:-5zvcxhprh">How to Use correctly:</h3>Use: `score_9`, `score_8_up`, `score_7_up`, `rating_explicit`, (if newds)Use Clip Skip 2. - Using the `danbooru` tags when prompting.<ul><li>Such as <code>anime_screencap</code> for 2d anime look in positive</li><li>Or <code>realism, 3d </code>for a realistic 2.5 look</li></ul>Additional keywords improve output such as: `realistic`, `masterpiece`, `scenery`, `beautiful eyes`, `best quality`, `very aesthetic`.- When prompting, Using less is better for output!!!- Recommended aspect ratios are `1024x1024`, `832x1216`, `1216x832`. Works flawlessly with many SDXL resolutions.- Highly Recommended to use SDFORGE (if lower VRAM) using a hires fix to txt2img outputs<ul><li>Hires.fix Upscaler: R-ESRGAN 4x+ Anime6B</li><li>Upscale 1.2-2</li><li>Hires Steps 20-25</li><li>Sampling Steps 25-40</li></ul><h3 id="difficulties:-ysxwh486p">Difficulties:</h3><ul><li>Sometimes generating multiple fingers and toes</li></ul><h3 id="mixes:-cso9op4mg">Mixes:</h3>TBA<h3 id="improvements-goals-fi2aov24n">Improvements Goals</h3><ul><li>better background</li><li>better prompting</li><li>less wording</li></ul>

Files with SHA256: 07f5c3d7ae3db8c44cf10bf63b10a0a34316e0f9bb393eaf858a26b38923659a