Trained locally using sd_scripts @ 512x512 resolution for 2000 steps. 25 image dataset trained for 80 epochs using Joycap for captions. Trained the T5XXL Text encoder as well as CLIP-L. Works best for me using the euler beta sampler in ComfyUI @ 0.8 strength with 2.7 - 3 Flux guidance.<a target="_blank" rel="ugc" href="https://instagram.com/luxlocosplay">https://instagram.com/luxlocosplay</a><a target="_blank" rel="ugc" href="https://www.patreon.com/luxlocosplay">https://www.patreon.com/luxlocosplay</a><a target="_blank" rel="ugc" href="https://twitter.com/luxlocosplay">https://twitter.com/luxlocosplay</a>

Luxlo [flux] v1

512x512, AdamW optimizer. Trained using both CLIP-L and T5XXL + full nlp captions.

Files with SHA256: 7068bffb9343c30e8a9b0d955d7407d7ee3212a86ecf2595753821b1bb4e0248