<p><strong>If you like my work (models/videos/etc.), feel free to contribute here:</strong></p><p><strong>Discord channel:</strong> <a target="_blank" rel="ugc" href="https://discord.gg/ZKX3Zym8FQ">https://discord.gg/ZKX3Zym8FQ</a></p><p><strong>Tips/Donations:</strong> <a target="_blank" rel="ugc" href="https://streamlabs.com/kasukanra/tip">https://streamlabs.com/kasukanra/tip</a></p><p></p><p><strong><u>Pastebin for prompts:</u></strong> <a target="_blank" rel="ugc" href="https://pastebin.com/ujzhEp1a">https://pastebin.com/ujzhEp1a</a></p><p></p><p>The LoRAs included are trained locally on the kohya_ss GUI repository created by <a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss">bmaltais</a> as well as on the Google colab notebook for kohya <a target="_blank" rel="ugc" href="https://colab.research.google.com/github/Linaqruf/kohya-trainer/blob/main/kohya-LoRA-dreambooth.ipynb">LoRA</a>.</p><p></p><p>The local version is trained on resolution 768 x 768 pixles while the colab notebook version is trained on resolution 512 x 512 pixels.</p><p></p><p>Images were produced using <strong>anything 4.5 pruned checkpoint</strong> in combination.</p><p></p><p>If possible, using the 768 pixel version as it was trained on a higher resolution!</p><p></p><p><strong>Make sure your downloaded LoRA name matches with the prompt. It may or may not be different for you. Check the console to see if the LoRA is found.</strong></p><p></p><h2 id="heading-44">Reproducibility</h2><p>If you want to <strong>reproduce</strong> the images below, I've prepared a more detailed version of the prompt settings and seed values in these <strong>pastebin(s)</strong>:</p><p></p><p>768 version: <a target="_blank" rel="ugc" href="https://pastebin.com/LwrqdYe3">https://pastebin.com/LwrqdYe3</a></p><p></p><p>512 version: <a target="_blank" rel="ugc" href="https://pastebin.com/wMa1gEJG">https://pastebin.com/wMa1gEJG</a></p><p></p><h1 id="heading-45">Stable Diffusion Workflow Videos</h1><p><strong><span style="color:rgb(255, 255, 255)">High-resolution advanced inpainting with ComfyUI (SDXL, IP-Adapter, rgthree)</span></strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/EhrArMjIDZw" start="0"></iframe></div><p><strong><span style="color:rgb(255, 255, 255)">Improving Stable Diffusion images with FreeU (SDXL, LCM, Turbo) and ComfyUI API</span></strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/WwsJ_QIgsG8" start="0"></iframe></div><p><strong><span style="color:rgb(255, 255, 255)">Autogen/AI Agents &amp; Local LLMs autonomously create realistic Stable Diffusion model</span></strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/jUUk4RQpiK4" start="0"></iframe></div><p><strong><span style="color:rgb(255, 255, 255)">Worldbuilding an IP with 3D and AI ft. MVDream | Part 1</span></strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/hIsvjfU5xO8" start="0"></iframe></div><p><strong>ControlNet 1.1 for Character Design | Part 2</strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/6bksNeiMP9M" start="0"></iframe></div><p></p><p><strong>Creating a PNGtuber with Character LoRA</strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/JgYpUjxRJR0" start="0"></iframe></div><p></p><p><strong>Training LyCORIS (LoCon/LoHa) for Style | LoRA Part 2</strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/Icf3ZQ67KPI" start="0"></iframe></div><p></p><p><strong>ControlNet for Character Design | Part 1</strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/Q8vg73MMeE4" start="0"></iframe></div><p></p><p><strong>Training LoRA for Style | LoRA Part 1</strong></p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/7m522D01mh0" start="0"></iframe></div><p></p><p><strong>Creating Stylized Vector Logo Characters</strong></p><div data-youtube-video><iframe width="100%" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/w-wMXn-EsGw" start="0"></iframe></div><p><strong>How I train DreamBooth for Style</strong>:</p><div data-youtube-video><iframe width="100%" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/R5VrS-JXuNg" start="0"></iframe></div><p></p><p><strong>Model Demonstration:</strong></p><div data-youtube-video><iframe width="100%" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/aNewHiib4Nk" start="0"></iframe></div><p></p>

swd3e2 style LoRA

<p>LoHa doesn't perform that well. I don't recommend it, but it's here for posterity.</p><p></p><p>Trained LoCon on resolution 768 x 768 pixel dataset.</p><p>14 dimensions for all (network dim, alpha, conv dim, conv alpha)</p><p>DAdaptation Optimizer</p><p>Cosine LR Schedule</p><p>10% warmup</p><p>TE LE = 0.5</p><p>LR LE = 1</p>

Files with SHA256: f6e39f762557d2292f86cd176c428e4af0d25283e92ea26bb6fdbe7ced81c75d