An anime model based on NoobAI XL.<h3 id="v4.5-rectified-flow">v4.5 Rectified Flow</h3>Many thanks to <a rel="ugc" href="https://civarchive.com/user/bluvoll">Bluvoll</a> and <a rel="ugc" href="https://huggingface.co/CabalResearch">CabalResearch</a>. Without their pioneering work on converting Noob vpred to RF, this model would not exist.Important This is a Rectified Flow model, so you must use the ModelSamplingSD3 node (or equivalent).What's new: Enhanced lighting along with more diversity Enhanced colour Enhanced and more consistent details (especially visible after Hi-res fix) Reduced blue biasOverall, if you liked the improvements vpred brought to you over eps, RF further improves upon vpred.Recommended settings:<ul><li>Shift 3.0</li><li>CFG 3.0 is a good starting point, do experiment</li><li>Euler (DPM++ 2M SDE is also an option for Hi-res fix)</li><li>Simple</li><li>30 steps</li></ul><h3 id="older-versions">Older versions</h3>vpred v4I decided to revisit Tiro because:<ul><li>I got a couple of feedback that helped me realise v3.1 is flawed</li><li>bluvoll's Noob RF exposed how much potential is left untapped in Noob</li></ul>After many training experiments, I have a model that I think is worth sharing.Updates in vpred v4:<ul><li>Reduced blue bias</li><li>Reduced exposure issues</li></ul>Recommended settings:<ul><li>CFG 3.0 - 5.0</li><li>DPM++ 2M SDE / Euler a</li><li>SGM Uniform</li><li>30 steps</li></ul>Alternative settings:<ul><li>Euler / DPM++ 2M SDE</li><li>Exponential</li><li>30 steps</li></ul>Notes:<ul><li>Reducing exposure issues enabled more variety in lighting, however this also means not all gens will have pleasing lighting, so you might have to roll your seed</li><li>For some prompts, model might assume low light/night; add tags like day, add tags like dark, dim lighting, night, chiaroscuro in the negative, add light sources like sunlight, lamp, window light, light bulb, light particles to brighten it up</li></ul>vpred v3.1Recommended settings:<ul><li>CFG 3.0 - 5.0</li><li>DPM++ 2M SDE / Euler a</li><li>SGM Uniform</li><li>30 steps</li></ul>Updates in vpred v3.1:<ul><li>Finetuned from base Noob vpred</li><li>Mitigates the overexposure issues without significantly impacting the dynamic range and saturation</li><li>Much better at flat and sketchy styles</li><li>Includes a finetuned VAE with subtle improvements</li></ul>Important:<ul><li>If the default style is not to your taste, use artist tags</li><li>The aim of this finetune was to specifically try mitigate the issues present in base Noob vpred; hence the model's knowledge remains similar</li><li>very awa is a strong tag, use it prudently</li></ul>Note: If you want apply the changes to your own Noob vpred based model, grab the LoRA extraction from my HuggingFace.vpred v3.0Recommended settings:<ul><li>CFG 3.0 - 5.0</li><li>DPM++ 2M SDE</li><li>SGM Uniform</li><li>30 steps</li></ul>Updates in vpred v3.0:<ul><li>Finetuned from base Noob vpred, with no other models merged into the UNet</li><li>Much better at flat and sketchy styles</li><li>Less overexposure issues</li><li>Better at dark compositions - use tags like dark, night</li><li>Alternate aesthetic tag: very aesthetic</li></ul>Important:<ul><li>If the default style is not to your taste, use artist tags</li><li>The aim of this finetune was to specifically try mitigate the issues present in base Noob vpred; hence the model's knowledge remains similar</li><li>If you want more compatibility / less bias, grab the base variant from my HuggingFace</li></ul>vpred v2.0Recommended settings:<ul><li>CFG 3.0</li><li>DPM++ 2M SDE</li><li>SGM Uniform</li><li>30 steps</li></ul>Updates in vpred v2.0:<ul><li>Better white balance</li><li>Less oversaturation issues</li><li>Reduced style bias</li></ul>Note: If you want even less style bias, grab the base variant from my HuggingFacevpred v1.0Important: vpred version requires much lower CFG than eps models. CFG 1.5 is a good starting point for vpred version. CFG 4.0 is a good starting point for eps versions. Reduce CFG if you find your gens oversaturated.Your choice of sampler is important for the vpred version. Use ddpm for balanced colours, use stronger samplers like the SDE ones for more saturation.Follow the same prompting and configuration guidelines as NoobAI XL, especially for vpred version. Use artist tags if the default style of the model is not to your taste.I would love to see your gens and creativity!Inherits the same license as NoobAI XL.

LS Tiro XL

Further trained with larger dataset and improved recipe

lsTiroXL_vpredV4.safetensors

vpred_v4

v4.5_RF

vpred_v3.1

vpred_v3.0

vpred_v2.0

vpred_v1.0

eps_v2.0

eps_v1.0

Files with SHA256: d851833b09c2202b526d40eb8c271e38a723f2be208ca412086792d139185e52