First attempt at a LoRA on CivitAI. Trained locally using sd_scripts @ 1024x1024 resolution. 20 image dataset trained for 70 epochs. Works best for me using the dpmpp_2s_ancestral beta sampler in ComfyUI @ 0.8 strength with 2.5 - 2.8 Flux guidance.<a target="_blank" rel="ugc" href="https://www.instagram.com/evieleemikomin/">https://www.instagram.com/evieleemikomin/</a><a target="_blank" rel="ugc" href="https://www.patreon.com/mikomin">https://www.patreon.com/mikomin</a><a target="_blank" rel="ugc" href="https://x.com/MikominCosplay">https://x.com/MikominCosplay</a>

Evie Lee Mikomin [flux] v1

Exact same settings as V1.0 trained on same dataset just @ 512x512 instead of 1024x1024. Somethings are better in this version but generally I prefer v1.0.

Files with SHA256: 2fa79d60dd7a3b39808283b67196c26e9c9d0ec3b05d52af34373bce8c0506f9