<h1 id="***please-read***-i2w234ixh">***PLEASE READ***</h1>This is a lora I trained to depict Hinata from part 1 of Naruto.The trigger word is "Hinata". After that I generally lead my prompts with "an anime scene of..."This lora was not meant to replicate the sharpness and high quality of modern anime/anime artwork. It was designed to create reasonably believable part one screenshots.All of the example images were generated at 1024x1024 using the rank16 version at 1.0 strength, 3.5 guidance, dpmpp2m, beta and 25 steps. Then upscaled by 1.5 at 0.55 denoise. I used fp16 Flux Dev.<h2 id="versions-waaa6kfcm">Versions</h2>I've posted three versions.Rank 128 - This is the full lora that I trained. It is the most accurate, at the cost of overfitting. It is not very flexible and will likely conflict with other loras. The main reason I posted it is to give you the option of downloading it and resizing it to whatever you want.Rank 16 (recommended) - I find this to be a good middle ground between accuracy and flexibility. If it seems to be conflicting with other loras, then you can either decrease strength or use Rank 8.Rank 8 - This is the most flexible version at the cost of character fidelity. Expect it to work well with other loras, but some character details might not be exactly right due to loss of information.Please let me know if you have an tips or suggestions and consider donating buzz if you enjoy my work.

Hinata Hyuga [Part 1]

***PLEASE READ***This is a lora I trained to depict Hinata from part 1 of Naruto.The trigger word is "Hinata". After that I generally lead my prompts with "an anime scene of..."This lora was not meant to replicate the sharpness and high quality of modern anime/anime artwork. It was designed to create reasonably believable part one screenshots.All of the example images were generated using the rank16 version at 1.0 strength, 3.5 guidance, dpmpp2m, beta and 25 steps. Then upscaled by 1.5 at 0.55 denoise.VersionsI've posted three versions.Rank 128 - This is the full lora that I trained. It is the most accurate, at the cost of overfitting. It is not very flexible and will likely conflict with other loras. The main reason I posted it is to give you the option of downloading it and resizing it to whatever you want.Rank 16 (recommended) - I find this to be a good middle ground between accuracy and flexibility. If it seems to be conflicting with other loras, then you can either decrease strength or use Rank 8.Rank 8 - This is the most flexible version at the cost of character fidelity. Expect it to work well with other loras, but some character details might not be exactly right due to loss of information.Please let me know if you have an tips or suggestions and consider donating buzz if you enjoy my work.

Files with SHA256: 4c41b1b5ca52dc15dfd6e8b8a6d4255ab8dab4e46924146003388e229f59209e