Trained on 40 images and captioned with the civitai lora system. Edited the the tags on all 40 images to ensure Wartortle and not Squirtle (which is how it was detected) was named. For v1, I ended up training overnight on my 3090 using ai-toolkit and the flux-dev2pro model (flux-dev fine tuned for more steps to train better).

Wartortle (Pokemon) [FLUX]

Sorry about this one, it got overbaked and civitai deleted all the other epochs instantly :( Training on my 3090 overnight for a proper V1

Files with SHA256: bce96ff9f74b5c010dd423e9a1612c2b1dc99cde323a7ddf4c9219b3c612fc06