Left: v1, right: v2.Everyone has a dream they strive to achieve, and so does the musically talented Kanon Shibuya. Consider adding those to the negative prompt if you don't want them to appear in your images. Might confuse left and right bangs if generate sideway imagesĪOM3A1 also has the tendency to change Kanon's hairstyle to hair bun, hair tie, or add random hair bow, ornament, etc. Good for generating anime style artĪOM3A1: around 0.6. Notes for v1Īnything v4.5: use a LoRA ratio of around 0.6-0.8Ĭorneos7thHeavenMix: ratio about 0.7-0.8. Also precisely describes Kanon, though generated images have high saturationĭemo images did not went through any highres upscaler. Do not get fooled by the name, it is actually a good SFW modelĪnything v4.5, AOMv3: similar to the above. I recommend using this since it is what the LoRA has been fine-tuned from.

See changelog ( About this version section on the left pane) for more details regarding the training parameters. The key to this might be using a high repeat number and low learning rate, as well as a large enough LoRA capacity (I used 32/16). Most importantly, it rarely distorts her eyes. This version produces images extremely close to the anime art style. For v2, include bangs, hair between eyes as well (the model will generate her wearing glasses without bangs if not prompted) To generate school uniform: student uniform, gray dress, red ribbon, neck ribbon, pinafore dress, white shirt This is important, as previous models I trained for Kanon as well as other characters lacked the ability to generate sideway images. 20-30 images are chosen from sideway angles (booru tag from side). Trained using around 150 anime screenshots from season 1 and season 2. The final checkpoint is created by averaging the checkpoints of epoch 12 and 8: 0.8*epoch_12 + 0.2*epoch_8. Pretrained_model_name_or_path = '/content/ashleymix-v2.safetensors' Logging_dir = '/content/drive/MyDrive/lora_training/log' Output_dir = '/content/drive/MyDrive/lora_training/output/shibukano' The model is trained with below parameters and repeat 8:

Inference should work on any model, but Meina_Hentai might give the best result (it is a good SFW model, don't get fooled by the name). The difference is that it is fine-tuned from my custom checkpoint with 0.7*Meina_Hentai + 0.3*AnyLoRA.

This new version is trained on the same dataset as before, with 151 anime screencaps.

Love live superstar download