<p>You can use two of my models for online generation on Tensor Art:<br /><a target="_blank" rel="ugc" href="https://tensor.art/models/849516334959261898">Illustrious version 3</a><br /><a target="_blank" rel="ugc" href="https://tensor.art/models/853746118608051226">NAI version</a><br /></p><h3 id="but-if-you-want-to-use-this-model-here-on-civit-then-consider-bidding-some-buzz-to-make-it-available-on-the-generator!i-will-bid-all-the-buzz-i-get-from-all-of-you-using-the-model!-(if-i-remember.)current-bid:-750-buzz-on-nai-v-pred">But if you want to use this model here on Civit, then consider bidding some <span style="color:rgb(250, 176, 5)">buzz</span> to make it available on the generator!<br /><span style="color:rgb(190, 75, 219)"><em>I will bid all the buzz I get from all of you using the model! (If I remember.)</em></span><br />Current bid:<span style="color:rgb(250, 176, 5)"> </span><span style="color:rgb(64, 192, 87)">750</span><span style="color:rgb(250, 176, 5)"> buzz </span>on<span style="color:rgb(250, 176, 5)"> </span><span style="color:rgb(250, 82, 82)">NAI V-Pred</span></h3><p></p><h1 id="noobai-version:">NoobAI version:</h1><p>I wanted to be part of the cool gang and hop on the NAI v-pred craze. Here's what I cooked up.<br /></p><p>As with any of these models, use Euler or Euler A samplers. Comfy user should use ModelSamplingDiscrete and set it to v-prediction. I'll add more info here as I continue to research but for now, resources.</p><p></p><p>You can find CSV files filled with tags the model understands <a target="_blank" rel="ugc" href="https://huggingface.co/datasets/Laxhar/noob-wiki/tree/main">here</a>. Both danbooru and e621. Which... just means prompting like you would an Illustrious model.<br /><br />Illustrious LoRA models can break contrast and saturation, but they should work. Please report any issues you find.<br /><br />ComfyUI users can use the example workflow <a target="_blank" rel="ugc" href="https://huggingface.co/Laxhar/noobai-XL-Vpred-0.5/blob/main/comfy_ui_workflow_sample.png">here</a>. Just drag the image into comfy to open the workflow.</p><p><a target="_blank" rel="ugc" href="https://civarchive.com/articles/8962/noobai-xl-quick-guide">General guide for using NoobAI models effectively.</a></p><p></p><h1 id="illustrious-version:">Illustrious version:</h1><p>Triple model merge. It's a first merge attempt, but it came out alright I'd say. Understands Danbooru and e621 tags.</p><p></p><p>It's an EPS model but it seems to work fine without EPS sampling, but I'd still recommend you use the correct settings for whichever tool you use. (ComfyUI/A1111/etc.)</p><p>I do not use A1111, but for ComfyUI you have to use the node ModelSamplingDiscrete or an equivalent custom node and set it to EPS sampling.</p><p></p><p>Use Euler A sampler, ~25-40 steps and ~4-8 guidance scale (CFG).</p><h2 id="positive-illustrious-prompts:"><span style="color:rgb(130, 201, 30)">Positive Illustrious prompts:</span></h2><pre><code>(subject, {1boy/1girl/2boys/2girls/etc.}, artist/style).

Describe image with natural language. {A portrait of ...}.

(tags, quality tags).</code></pre><h2 id="negative-illustrious-prompts:"><span style="color:rgb(250, 82, 82)">Negative Illustrious prompts:</span></h2><pre><code>bad quality, worst quality, worst detail, sketch, censor.</code></pre><p></p><h1 id="pony-version:">Pony version:</h1><p>Test model, I wanted to see if merging two of my favorite models would increase prompt adherence while still keeping its imagination.</p><p></p><p>It has a bit of a watermark problem, but they don't seem to be <em>too</em> intrusive and are usually easily fixed with a bit of inpainting.</p><p></p><h2 id="positive-pony-prompts:"><span style="color:rgb(130, 201, 30)">Positive Pony prompts:</span></h2><pre><code>score_9, score_8_up, score_7_up, rating_tag, BREAK</code></pre><h2 id="negative-pony-prompts:"><span style="color:rgb(250, 82, 82)">Negative Pony prompts:</span></h2><pre><code>score_6, score_5, score_4, etc.</code></pre><h3 id="but-please-experiment-i'd-love-to-hear-your-thoughts!">But please experiment, I'd love to hear your thoughts!</h3><p></p>

IndigoDreamXL

<p>This model is practically identical to v1 in style, some small tweaks here and there, but what this version primarily improves is shapes and anatomy. I will be honest and say it's pretty hard to tell, even when they're side by side in my comparisons. Prompt structure is the same. This one has VAE built in, but you're not required to use it.</p><p>DARE-TIES merge &lt;<a rel="ugc" href="https://civitai.com/models/1015206?modelVersionId=1342462">IndigoDreamIL </a>- <a rel="ugc" href="https://civitai.com/models/1017790?modelVersionId=1141578">Fifth_Mix</a>&gt;</p><pre><code>drop_rate: 0.8

# Block Gradient
time: 1
label: 1
input: 0.2
middle: 0.7
output: 0.5
out: 0.5</code></pre><p>2nd DARE-TIES merge &gt; <a rel="ugc" href="https://civitai.com/models/1091817?modelVersionId=1352516">WesternToons-IL</a></p><pre><code>drop_rate: 0.9

# Block Gradient
time: 1
label: 1
input: 0.8
middle: 1
output: 0.8
out: 0.7</code></pre>

Files with SHA256: f6efdf01d72eb4d8666458be144ce53f1a24219b02b339e5aa0fddc49f86863a