<h3 id="to-update-the-model-i-have-spent-quite-some-buzz-to-generate-loras."><span style="color:rgb(21, 170, 191)">To update the model, I have spent quite some buzz to generate LORAs.</span></h3><h3 id="hence-the-early-access-the-buzz-will-be-used-to-update-newer-versions."><span style="color:rgb(21, 170, 191)">Hence the early access, the buzz will be used to update newer versions.</span></h3><p></p><p><em>If you enjoy using the model, I appreciate if you can do any of the following:</em></p><ul><li><p><strong>Upload an image and tag the checkpoint</strong>. (I get buzz from each unique user that posts an image to model so even one upload helps alot. I also love to see uploads made with the model)</p></li><li><p><strong>Send a tip</strong> (even 10 buzz is appreciated)</p></li><li><p><strong>React to the latest images</strong> I have uploaded.</p></li></ul><p></p><h3 id="thank-you-for-the-support.">Thank you for the support.</h3><hr /><h1 id="version-notes"><span style="color:rgb(250, 82, 82)">Version notes</span></h1><h3 id="v2">v2</h3><ul><li><p>Increased colour saturation, brightness.</p></li><li><p>Increased head size/reduced broad shoulders.</p></li><li><p>Less dreamy, lesser artifact.</p></li><li><p>Prompting should be more responsive.</p></li></ul><h3 id="v1">v1</h3><ul><li><p>First version that is on par/better than the illustrious mix</p></li></ul><hr /><h1 id="version-features"><span style="color:rgb(250, 82, 82)">Version features</span></h1><h3 id="rather-than-improvements-each-version-has-their-own-strengths"><em>Rather than improvements, each version has their own strengths</em></h3><ul><li><p>v1 - Dark, desaturated, dreamy style</p></li><li><p>v2 - Bright, natural style</p></li></ul><hr /><h1 id="model-focus"><span style="color:rgb(250, 82, 82)">Model focus</span></h1><ul><li><p>3DCG/Semi realistic face; sharper/mature facial features with subtle skin texture.</p></li></ul><h3 id="weakness">Weakness</h3><ul><li><p>Noise pattern on image. Recommend using WAN 2.1 fp32 VAE and film grain postprocessing node.</p></li><li><p>Hands can be funky sometimes.</p></li><li><p>Anime style hairstyle can be hard to prompt</p></li></ul><hr /><h1 id="recommended-settings"><span style="color:rgb(130, 201, 30)">Recommended settings</span></h1><h3 id="(can-see-workflow-via-uploaded-images)"><span style="color:rgb(130, 201, 30)"><strong>(can see workflow via uploaded images)</strong></span></h3><h3></h3><p><strong>euler + simple</strong></p><ul><li><p>steps 8-10, cfg 1</p></li><li><p>standard, simple, stable, works most of the time</p></li><li><p>downside - lacks texture, uniform texture, softer looking</p></li></ul><p></p><p><strong>euler + beta/(beta57-more tex)</strong></p><ul><li><p>steps 8-12, cfg 1</p></li><li><p>more texture than simple but can cause more artifacts</p></li></ul><p></p><p><strong>euler_a + beta57</strong></p><ul><li><p>steps 8-15, cfg 1</p></li><li><p>even more texture and details but also can get the image wrong</p></li><li><p>can cause the image to look more illustrative/less realistic</p></li></ul><p></p><p><strong>deis_2m / deis_3m + beta57</strong></p><ul><li><p>steps 3-4, cfg 1</p></li><li><p>similar to euler+beta57 but can have more stable structure while having as much texture</p></li><li><p>might have more realistic proportions compared to euler</p></li><li><p>downside - can be soft?</p></li><li><p>deis_3m is longer but might have better structure</p></li></ul><hr /><h1 id="remarks"><span style="color:rgb(250, 82, 82)">Remarks</span></h1><ul><li><p>Personally use 1024x1536 resolution most of the time. Lower resolutions makes the noise pattern more noticeable when upscaled.</p></li><li><p>Upscale model recommendation: <a rel="ugc" href="https://openmodeldb.info/models/2x-LiveActionV1-SPAN">2xLiveActionV1_SPAN</a>. Keeps the image similar to the original without extra sharpening or lost of details while 2x resolution. Also fast.</p></li><li><p><a rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_fp32.safetensors">WAN 2.1 fp32 VAE</a> and film grain postprocessing node for noise pattern reduction.</p></li></ul><p></p>

FUBUWA MIX 3D REAL KREA2

<p>v2</p><ul><li><p>Increased colour saturation, brightness.</p></li><li><p>Increased head size/reduced broad shoulders.</p></li><li><p>Less dreamy, lesser artifact.</p></li><li><p>Prompting should be more responsive.</p></li></ul>

Files with SHA256: 5d322dd6f219be4ac6520b98669f27859624ec4d69410fb5d23c76bbb6ef74c5