<h1 id="mitsuartxl-style-lora-lw2fg1m6k">MitsuArtXL Style LoRA</h1>A versatile LoRA that maintains high image quality and consistency while playing well with other styles. Recommended weight: 0.6 (full weight can cause random hair color shifts).<h2 id="strengths-3yol17hns">Strengths</h2><ul><li>Decent hand rendering, particularly with ADetailer</li><li>Strong anatomy consistency across human and non-human subjects</li><li>Handles diverse skin tones well (tested with light, dark, and red skin)</li><li>Works effectively with animal subjects and furry content</li><li>Good weapon rendering, even with complex prompts</li><li>Maintains quality with wide-shot compositions despite slight edge blur</li><li>Compatible with other art styles (tested with Kajin and Naga U)</li><li>Handles chibi style adequately</li><li>Capable of extreme poses while maintaining reasonable anatomy</li><li>Drop shadow effects render properly</li></ul><h2 id="characteristics-z4xqajox7">Characteristics</h2><ul><li>Tends toward bright lighting and vibrant colors</li><li>Best results at 0.6 weight (full weight can cause hair color inconsistency)</li><li>Slight blur effect in wide-shot compositions</li></ul><h2 id="limitations-1yv7yx3ds">Limitations</h2><ul><li>Struggles with expressive faces (&gt;_&lt;, DX expressions) unless mixed with other styles</li><li>Some specific poses may be difficult to achieve</li><li>Can have occasional anatomical inconsistencies in certain image areas</li><li>May need style mixing for certain specialized expressions</li></ul><h2 id="technical-notes-xib84cnst">Technical Notes</h2><ul><li>Significantly outperforms epoch 81, particularly in fine detail rendering</li><li>More reliable ADetailer performance compared to epoch 81</li><li>Testing shows consistent performance across various subject matter and compositions</li></ul><h2 id="model-details-twzzbnji8">Model Details</h2>SDXL Version:<ul><li>Weight: 0.8 - 1.0 (Testing shows 0.6 optimal for epoch 91)</li><li>Trigger: <code>mitsuartstyle</code></li></ul>SD 1.5 Version:<ul><li>Weight: 0.8 - 1.0</li><li>Trigger: None required</li></ul><h2 id="technical-training-details-1dxr19g3n">Technical Training Details</h2>Base Resolution: 1024 Network Dim: 32 Network Alpha: 16 Optimizer: Adafactor Scheduler: Cosine with Restarts Training Steps: 2525 Epochs: 100 Batch Size: 8<pre><code>{
 "engine": "kohya",
 "unetLR": 0.0001,
 "clipSkip": 1,
 "loraType": "lora",
 "keepTokens": 1,
 "networkDim": 32,
 "numRepeats": 1,
 "resolution": 1024,
 "lrScheduler": "cosine_with_restarts",
 "minSnrGamma": 5,
 "noiseOffset": 0.1,
 "targetSteps": 2525,
 "enableBucket": true,
 "networkAlpha": 16,
 "optimizerType": "Adafactor",
 "textEncoderLR": 0.00005,
 "maxTrainEpochs": 100,
 "shuffleCaption": true,
 "trainBatchSize": 8,
 "flipAugmentation": true,
 "lrSchedulerNumCycles": 3
}</code></pre><h3 id="tips-i0ldxw3xe">💡 Tips</h3><ul><li>Start with recommended weights (0.6 for best results with epoch 91)</li><li>Adjust based on your specific needs</li><li>Works well with both simple and complex prompts</li><li>Excellent for style mixing and combinations</li></ul><h2 id="credit-7rqhg2klk">📜 Credit</h2>Based on the artistic style of MitsuArt. Please respect the original artist's work when using this LoRA.If you find this LoRA useful, don't forget to leave a like and share your creations! Feedback and suggestions are always welcome.

MitsuArt Style

Trained LoRA to capture Mitsuart's art style on SDXL. Built with 134 curated images over 100 epochs.Works best at 1.0 strength, but can use 0.6 - 0.8 for subtle influence. Trigger: "mitsuartstyle"Training:- 134 images × 100 epochs- Dim/Conv: 32, Alpha: 16- AdaFactor+ Huber loss- Cosine w/restarts scheduler- LR: 0.0001 + Unet LR: 0.00005- Dropouts: 0.08- Base: NoobAI Epsilon 1.0Enjoy! 🎨

Files with SHA256: 4b8b878df3973e887fb0cb210a719f03962d03d7263c196c0f100e0caee3edd6