AkashicPulse EQVAE is a model based on AkashicPulse series, but fine-tuned to utilize <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE">Anzhc's EQVAE</a> as its VAE. (In theory,) This improve the model generation color and accuracy.Recommended settings [UPDATED FOR V1.0.x]:<span style="background-color:rgb(26, 27, 30);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">Same like v0.2, the model is in between needing RescaleCFG and not needing it. Expect to turn on and off this a lot.<ul><li>The usage of artist tags IS HIGHLY RECOMMENDED. Put it at the FRONT of your prompt (and use BREAK if your generation platform support it).</li><li>Steps: 32-50</li><li>CFG: 7-10 with RescaleCFG on or off, or 4.5-6 if you don't wanna touch RescaleCFG</li><li>RescaleCFG: 0.7 OR OFF, inspect generated image quality if something is off.</li><li>Sampler: Euler a (or try <a target="_blank" rel="ugc" href="https://github.com/nawka12/adept-sampler">AkashicSolver available on Adept Sampler</a> with AkashicAOS).</li><li>Noise Scale: 0.9, using 1.0 will cause the model to over-generate noise.</li></ul>Recommended negative prompt:<pre><code>worst quality, low quality, bad anatomy, bad hands, text, signature, watermark, username, artist name, [things you want to remove]</code></pre>WARNING: Using excessive negative prompt might resulting the model to lose its quality.Model Specs:v1.0.2<ul><li>Based on: AkashicPulse EQVAE v1.0.1-Base</li><li>Dataset size: ~21k</li><li>Epoch posted: Epoch 10 out of 10</li><li>EQVAE used: <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE/resolve/main/EQB7.safetensors">Anzhc's EQVAE B7</a></li></ul>v1.0-v1.0.1<ul><li>Based on: AkashicPulse EQVAE v0.2</li><li>Dataset size: ~4.6k</li><li>Epoch posted: Epoch 15 out of 15</li><li>EQVAE used: <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE/resolve/main/EQB7.safetensors">Anzhc's EQVAE B7</a></li></ul>v0.2 [CHECK ABOUT THIS VERSION FOR DETAILS]<ul><li>Based on: AkashicPulse EQVAE v0.1</li><li>Dataset size: ~9.4k</li><li>Epoch posted: Epoch 7 out of 10.</li><li>EQVAE used: <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE/resolve/main/EQB7.safetensors">Anzhc's EQVAE B7</a></li></ul>v0.1<ul><li>Based on: <a target="_blank" rel="ugc" href="https://civarchive.com/models/1055452?modelVersionId=2426269">AkashicPulse 4.1</a></li><li>Dataset size: 10.6k, repeated twice.</li><li>Epoch posted: Epoch 9 out of 10.</li><li>EQVAE used: <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE/resolve/main/EQB7.safetensors">Anzhc's EQVAE B7</a></li></ul>

AkashicPulse EQVAE

The v1.0.1 version I uploaded was a merge between this model and v1.0, which upon future testing feels keeping both bad side of v1.0 and this model.I'm not a merge master (despite starting my whole career here as a model merger), so maybe if I release this model someone will try merge and create something better.This model weakness:<ul><li>Less aesthetic than v1.0</li><li>Anatomy sometimes broke</li></ul>v1.0 weakness:<ul><li>Poor on white background</li><li>Needing much parameter settings to generate good images</li></ul>This model fixes v1.0 weakness while having new issue.

akashicpulseEQVAE_v101Base.safetensors

v1.0.1-Base

v1.0.4

v1.0.3

v1.0.2

v1.0.1

v1.0

v0.2

v0.1

[WARNING: This model is EXPERIMENTAL and not stable.]AkashicPulse EQVAE is a model based on AkashicPulse series, but fine-tuned to utilize <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE">Anzhc's EQVAE</a> as its VAE. (In theory,) This improve the model generation color and accuracy.Model Specs:v0.2 [CHECK ABOUT MODEL FOR DETAILS]<ul><li>Based on: AkashicPulse EQVAE v0.1</li><li>Dataset size: ~9.4k</li><li>Epoch posted: Epoch 7 out of 10.</li><li>EQVAE used: <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE/resolve/main/EQB7.safetensors">Anzhc's EQVAE B7</a></li></ul>v0.1<ul><li>Based on: <a target="_blank" rel="ugc" href="https://civitai.com/models/1055452?modelVersionId=2426269">AkashicPulse 4.1</a></li><li>Dataset size: 10.6k, repeated twice.</li><li>Epoch posted: Epoch 9 out of 10.</li><li>EQVAE used: <a target="_blank" rel="ugc" href="https://huggingface.co/Anzhc/MS-LC-EQ-D-VR_VAE/resolve/main/EQB7.safetensors">Anzhc's EQVAE B7</a></li></ul>Recommended settings [UPDATED]:<ul><li>The usage of artist tags IS HIGHLY RECOMMENDED. Put it at the FRONT of your prompt (and use BREAK if your generation platform support it).</li><li>Steps: 45</li><li>CFG: 7</li><li>RescaleCFG: 0.7 (Without this, expect the model will sometimes produce burnt images)</li><li>Sampler: Euler a (or try <a target="_blank" rel="ugc" href="https://github.com/nawka12/adept-sampler">AkashicSolver available on Adept Sampler</a> with AkashicAOS).</li></ul>WARNING: Due to dataset and training resources limitations, the model might break (e.g generation looks burned, bad anatomies, etc) under some random conditions. If this happens, try to change your prompt and negatives, steps, scheduler, or sampler.

Files with SHA256: f7b4de1b85e0e70dc425a77f14d40904e2b5d4287bb3c7de04ecca095958d93f