Stable Diffusion

Ether Blu Mix Ether Blu Mix 5

model_id: etherblumix-etherblumix5

<p><strong><em>* Version 5 Update:</em></strong></p><p>Ether Blu Mix 5 looks to further enhance the vibrant, painterly look with my first attempt at LORA mixing. There's also a lot more <a target="_blank" rel="ugc" href="https://civitai.com/models/4468/counterfeit-v30"><em>Counterfeit 3</em></a><em> </em>allowing for more compositional variety.</p><p>EBM 4 did a lot of things right; as a base it created detailed, vibrant images reminiscent of a classic anime style thanks to <a target="_blank" rel="ugc" href="https://civitai.com/models/12262/night-sky-yozora-style-model"><em>YOZORA</em></a>. Though, personally, it seemed too biased at depicting characters at a somewhat close distance, typically creating waist-high portraits. Adding more weight to Counterfeit 3 should hopefully add more compositional variety, while slightly reducing <em>YOZORA</em> and <a target="_blank" rel="ugc" href="https://civitai.com/models/21200/color-box-model"><em>Color Box's</em></a> weights to maintain EBM 4's style. <a target="_blank" rel="ugc" href="https://huggingface.co/nuigurumi/basil_mix"><em>Basil Mix</em></a> was added as a supplementary realistic model in tandem with <a target="_blank" rel="ugc" href="https://civitai.com/models/4550/ayonimix"><em>AyoniMix</em></a>. It's a personal favorite that plays nice with anime (booru tag) style models.</p><p>EBM 5 is my first attempt at LORA mixing. I was initially hesitant at merging LORA's to checkpoints as they might introduce unwanted bias, since LORA's tend to be trained on small datasets. After plenty of testing, as well as learning from Luna's documentation of <a target="_blank" rel="ugc" href="https://civitai.com/models/3850/kenshi"><em>Kenshi</em></a> and <a target="_blank" rel="ugc" href="https://civitai.com/models/40199/aurora"><em>Aurora</em></a>, I applied a light mix of LORA's to help enhance the vibrant, painterly style, while also improving background expression.</p><p><strong><em>Your prompts might react differently as the base of the model is no longer just AOM3, so please experiment and have fun!</em></strong></p><p><strong>📘 <em>Model Merging Info - v5</em></strong></p><p>I have no training or real knowledge of machine learning. I merged the model using the information provided on <a target="_blank" rel="ugc" href="https://huggingface.co/WarriorMama777/OrangeMixs"><em>WarriorMama777's AOM model page.</em></a> I very arbitrarily used the UBM weights given to merge and create AOM3 and its derivatives and modified the values with trial and error until I was visually happy with it.</p><p>I utilized the <a target="_blank" rel="ugc" href="https://github.com/hako-mikan/sd-webui-supermerger"><em>SuperMerger</em></a> extension for AUTO1111 by hako-mikan.</p><p><em>▼</em><strong><em> Merging</em></strong></p><ul><li><p>MODEL A: AOM3.safetensors</p></li><li><p>MODEL B: Counterfeit-V3.0_fp16.safetensors</p></li></ul><p>Add Sum merge - <em>Don't Use MBW</em></p><ul><li><p>Alpha: 0.5</p><ul><li><p>Model: AOM3-Counter3Sum-05</p></li></ul></li></ul><p>▼</p><ul><li><p>MODEL A: AOM3-Counter3Sum-05</p></li><li><p>MODEL B: nightSkyYOZORAStyle_yozoraV1PurnedFp16</p></li></ul><p>Add Sum merge - <em>Don't Use MBW</em></p><ul><li><p>Alpha: 0.4</p><ul><li><p>Model: AOM3-Counter3Sum-05-YozoraSum-04</p></li></ul></li></ul><p>▼</p><ul><li><p>MODEL A: AOM3-Counter3Sum-05-YozoraSum-04</p></li><li><p>MODEL B: colorBoxModel_colorBOX.safetensors</p></li><li><p>MODEL C: AOM3-Counter3Sum-05</p></li></ul><p>Add Difference merge - <em>Don't Use MBW</em></p><ul><li><p>Alpha: 0.2</p><ul><li><p>Model: AOM3-Counter3Sum-05-YozoraSum-04-ColorBoxDiff-02</p></li></ul></li></ul><p>▼</p><ul><li><p>MODEL A: AOM3-Counter3Sum-05-YozoraSum-04-ColorBoxDiff-02</p></li><li><p>MODEL B: pastelmix-better-vae-fp16.safetensors</p></li><li><p>MODEL C: Basil_mix_fixed.safetensors</p></li></ul><p>Weight Sum merge - <em>Use MBW</em></p><ul><li><p>Alpha: 1</p><p>0,0.2,0.2,0.2,0.2,0.2,0,0,0,0,0.2,0.1,0.2,0.2,0.2,0.2,0.2,0.1,0.1,0.1,0.2,0.2,0.1,0.2,0.2,0.2</p></li><li><p>Beta: 1</p><p>0,0.2,0.2,0.2,0.2,0.2,0,0,0,0,0.2,0.1,0.2,0.2,0.2,0.2,0.2,0.1,0.1,0.1,0.2,0.2,0.1,0.2,0.2,0.2</p><ul><li><p>Model: AOM3-Counter3Sum-05-YozoraSum-04-ColorBoxDiff-02-PastelSumS1-02-BasilSumS1-02</p></li></ul></li></ul><p>▼</p><ul><li><p>MODEL A: AOM3-Counter3Sum-05-YozoraSum-04-ColorBoxDiff-02-PastelSumS1-02-BasilSumS1-02</p></li><li><p>MODEL B: ayonimix_V5.safetensors</p></li></ul><p>Weight Sum merge - <em>Use MBW</em></p><ul><li><p>Alpha: 1</p><p>0,0.2,0.2,0.2,0.2,0.2,0,0,0,0,0.2,0.1,0.2,0.2,0.2,0.2,0.2,0.1,0.1,0.1,0.2,0.2,0.1,0.2,0.2,0.2</p><ul><li><p>Final Model: EtherBluMix-5</p></li></ul></li></ul><p>▼ <strong><em>LORA Mix</em></strong></p><ul><li><p>xlimo768: 0.25</p></li><li><p>dpep4-768-000004: 0.3</p></li><li><p>sttabi_v1.4-04: 0.05</p></li><li><p>Jordan_3: 0.2</p></li><li><p>lightingv2: 0.4</p></li></ul><p><strong><em>Models Used:</em></strong></p><pre><code>*AOM3.safetensors D124FC18F0232D7F0A2A70358CDB1288AF9E1EE8596200F50F0936BE59514F6D *Counterfeit-V3.0_fp16.safetensors CBFBA64E662370F59D4AA2AA69BF16749FCE93846CCCE20506AEE5DF01169859 *pastelmix-better-vae-fp16.safetensors D01A68AE76F97506363F387F5F28BB564AD9E20924844FD5945E600B72D39E79 *colorBoxModel_colorBOX.safetensors 93A20525F51798309AF22AC6E8285E61CF563A88F1EE00AA9F544CA939E63181 *nightSkyYOZORAStyle_yozoraV1PurnedFp16 E7BF829CFFD9EA034DA3B59FA13421F52AA4E4ED74FCBBEEFBF9CFB2F97C329E *ayonimix_V5.safetensors B46B04432B070FD88D69161B62C43EDCD4340DFC27FFB2E5FE7EDA64950C7E34 *Basil_mix_fixed.safetensors 0FF127093F5BE455057742C40CEF578407B6933F240EE8DC5ED0F3061196FB38 </code></pre>