Type | |
Stats | 305 |
Reviews | (17) |
Published | Aug 4, 2023 |
Base Model | |
Training | Epochs: 10 |
Usage Tips | Clip Skip: 2 |
Trigger Words | anizhur |
Hash | AutoV2 AE05D5A082 |
Anizhur (real name Anastasiia Zhurbenko) is a Russian model who makes video reviews of bikinis and swimsuits, from Mexico, where she lives. She is also a Instagram and OnlyFans model and cosplayer.
🔹If you can, add a drop to my ☕Ko-fi, here
🔹It will help me improving my lame hardware. 💖 Thanks!
🔹If you try it, add your review and share your results, I wanna see it!
As you can see from my tests, it works fine for both realistic or cartoon-style checkpoints. It seems to blend well with other LoRas, too.
📖 Prompt
Other than "anizhur", I strongly suggest you to try many other words on your prompt, because the training included multiple styles of clothes and poses (151 images were used).
Some tips: "barefoot, one-piece swimsuit, bracelet, jewelry, head tilt, long hair, blonde, tank top, dress, necklace, navel, pants, midriff, wristwatch, high heels, lying, kneeling, on stomach (lying), one-piece swimsuit, arched back, standing, jeans, feet, selfie", etc, etc.
I'm still finding what I can get from it with different checkpoints. But it's a lot.
🛠️ Resources used
Some resources used in my examples.
Note: Easy Diffusion used for all images, so my prompts are for Easy Diffusion (slightly different of Automatic1111, like the strength outside (), as "(anizhur)1.3" instead of "(anizhur:1.3)".
First example: wearing hat, bar background
• absolutereality_v16
• vae-ft-mse-840000-ema-pruned
• 448 x 640
• anizhur-v1 at 0.9
Second example: white dress portrait
• photon_v1
• vae-ft-mse-840000-ema-pruned
• 448 x 640
• anizhur-v1 at 1.0
Third example: cartoonish at night
• realcartoon3d_v5
• 576 x 704
• anizhur 0.76 and add_detail 0.8
Fourth example: military realistic
• photon_v1
• vae-ft-mse-840000-ema-pruned
• 512 x 640
• anizhur-v1 at 0.96
Fifth example: cute cyborg
• revAnimated_v122
• vae-ft-mse-840000-ema-pruned
• 448 x 640
• anizhur-v1 at 0.96 and nijiMecha at 0.5
Remember: this LoRa is based in a real person, so be respectful and smart about what you use it for. I'm not responsible for what you do with this LoRa.