Epsilon <a target="_blank" rel="ugc" href="https://files.catbox.moe/f41t0a.mecha">(recipe)</a> - <a target="_blank" rel="ugc" href="https://en.wikipedia.org/wiki/Orthogonal_Procrustes_problem">Rotate</a> <a target="_blank" rel="ugc" href="https://civarchive.com/models/895164/noobai-epsilon-05075-rotate-itercomp">the previous model</a> with <a target="_blank" rel="ugc" href="https://civarchive.com/models/246299?modelVersionId=997881">comradeship v14T13A</a> because both use 0.5 as a base.V-prediction <a target="_blank" rel="ugc" href="https://files.catbox.moe/yoxgpj.mecha">(recipe)</a> - Added this epsilon merge on a <a target="_blank" rel="ugc" href="https://civarchive.com/models/870578/noobai-vpred-itercomp-test">v-prediction itercomp merge</a> using perpendicular component.The tool used is <a target="_blank" rel="ugc" href="https://github.com/ljleb/sd-mecha">sd_mecha</a>.<hr />You can use these merges with LoRAs trained on their base, for improved results.<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/833294?modelVersionId=968495">Epsilon 0.5</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/833294?modelVersionId=962003">V-prediction test</a></li></ul>LoRA settings I use can be found here: <a target="_blank" rel="ugc" href="https://huggingface.co/reoe/lora/blob/main/epsilon-settings.toml">epsilon</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/reoe/lora/blob/main/v-prediction-settings.toml">v-prediction</a>(Settings are compatible with <a target="_blank" rel="ugc" href="https://github.com/derrian-distro/LoRA_Easy_Training_Scripts">this UI</a>)(You have to train the LoRA on the base model, not these merges)<hr />Recommended method for high-res:Scaling: 1.5x (<a target="_blank" rel="ugc" href="https://github.com/Ttl/ComfyUi_NNLatentUpscale">optional latent upscaler</a>)Model: <a target="_blank" rel="ugc" href="https://openmodeldb.info/models/4x-NomosUni-span-multijpg">4x-NomosUni-span-multijpg</a>0.5 strength with this <a target="_blank" rel="ugc" href="https://huggingface.co/deepghs/booru_yolo/tree/main/yolov8s_aa09">yolov8 model</a>I recommend using a CFG around 4 or 5.

Complex NoobAI Merge

V-prediction with Zero Terminal SNR sampling.

Files with SHA256: d98c40363cb33eb108d826b944bb9bd5a6fc98ec925331b73c2d87437eedd3ce