Type | |
Stats | 164 150 |
Reviews | (40) |
Published | Dec 25, 2024 |
Base Model | |
Training | Steps: 1,932 Epochs: 7 |
Usage Tips | Clip Skip: 2 Strength: 1 |
Trigger Words | sarada uchiha |
Hash | AutoV2 A54CD1DE8B |
Sarada Uchiha Blue Vortex v1
Story
12/25/2024 Pony Update: 🎠Released Pony version of this model in beta. Didn't have much time to test, so feedback is welcome! Post your pics!! Happy holidays 🎅Follow me on Bluesky @burnthevillage.bsky.social: https://bsky.app/profile/burnthevillage.bsky.social
Hi friend!
Introducing my best Lora yet, Sarada Uchiha! This character lora is based on her look Three years after the original Boruto series. Since the anime is probably years away, I heavily relied on manga monochrome illustrations, any official art released and some fan art (+photoshopped / AI Generated pics). The results have been phenomenal as I'm finally getting a good workflow for SDXL training (I find SD1.5 incredibly difficult in comparison :'[ )! Though she mainly wheres one outfit, (That I know of as I have only finished the original Boruto series and have not read the blue vortex manga) I trained this lora with 5 main concepts: Her regular outfit, Her Mangekyō Sharingan (Not sure if she actually has one / how she got one?), The obligatory Florida outfit ;), A pajamas outfit based on her pajamas from the original Boruto series and I threw in some swimsuit pics for good measure. I omitted NSFW pics from the dataset, so this lora might struggle with that a little (Still gets the job done). I recommend using this lora: POV All In One SDXL, as it eliminates the need for control-net or additional loras if you're into that... Other than that, I could potentially add her O.G. outfit in the future(?) or any outfit you guys recommend (if it makes sense). Leave a review and let me know how you would improve this lora!
Key Words:
Trigger words: sarada uchiha, woman,
Mangekyō Sharingan: symbol-shaped pupils, mangekyo eyes, red eyes, eye focus, close-up,
Default outfit: glasses, earrings, o-ring choker, black jacket, konohagakure symbol, black shorts, belt, thigh holster, thigh strap, footwear, high-heel sandals, toes,
Florida Outfit: florida outfit, alternate costume, glasses, earrings, o-ring choker, sun hat, hat flower, black romper, frills,
Pajama Outfit: glasses, earrings, o-ring choker, black crop top, black shorts, midriff, navel,
Framing shot: hand on hip, fully body, upper body, close-up,
Suggested settings:
As usual, I want the pictures to come off as natural as possible. This lora does decent eye illustrations, in fact using adetailer on the eyes will overwrite the Sharingan if that's what you're going for. This eliminates the need for tools like ADetailer, Controlnet, face restore or Inpainting. Examples shown are typically just what you would get from putting in text and using a high-res fix, which I suggest a 4x upscaler.
I had CLIP skip 2 on all pics, I trained the LORA with Clip Skip 2, full bf16, no baked in VAE.
The Sharingan loses detail the further the camera is away from her face and I've found decent success with "close-up" and "eye focus" to frame the shot.
I started using Forge Webui A fork of AUTOMATIC1111. It comes with new samplers and I used Euler a Turbo, but regular Euler a should be fine I just prefer the turbo.
I only did testing with AAM - AnyLoRA Anime Mix, but I trained with Animagine XL V3.1 as a base so I'm hoping it should work with most models like my last few loras. Please feel free to leave any feedback!
P.S. No blue vortex spoilers please!
Have Fun,
Florida-Man
Official HF repository: (Soon)
Your contribution is greatly appreciated and helps me to create more content. If you like my model, please consider donating: https://ko-fi.com/904floridaman
I'm open to commissions.