<h3 id="this-lora-model-is-trained-on-pony-realism.-2yionrryr">This lora model is trained on Pony Realism.</h3>If you use animation-style checkpoints, the resulting image can also be animatedWell... The training seemed harder than I thought.I trained it three times with different parameters. This one seems to work slightly better.<s>Actually, it didn't have the effect I wanted.</s>This is the image in the training set.↓<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/3a3c7ab8-ddd0-4b13-8dc5-a7fe22627d7a/width=525/3a3c7ab8-ddd0-4b13-8dc5-a7fe22627d7a.jpeg" />This is the image generated by lora. ↓<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/18f4596b-5fa2-42ea-b7b0-9da8f58e2219/width=525/18f4596b-5fa2-42ea-b7b0-9da8f58e2219.jpeg" /><s>This face doesn't look like it.</s>Maybe need to increase num_repeats.(This lora is 15).Or you may need to train the portraits separately.There may be a new version in the future.But that's it for now.<s>There is a cost to training on colab, after all.</s>

Files with SHA256: e461bfbb15774f3d64b9e50ce5b1a2ddc1039ce200272ea7e3f0cde7d7950caf