🧩Version Description

The Vpred version performs well with both the NoobAI Vpred and EPS checkpoints, and fine-tuned models are still recommended.

What’s the difference?

Images generated with Vpred feature a richer hand-painted texture, better color expression, and far away from sleek artificial rendering. This also brings better eye color rendering for the character Ai Hoshino. Note that you should match it with fine-tuned models that feature a hand-painted style.

🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺🔺

⚖️ Weight:    1

🔑 Trigger word:    hoshinoai

🎨 Outfit

hoshinoai, 1girl, star-shaped pupils, rabbit hair ornament, pink hair ribbon, pink frilled dress, heart brooch, blue star brooch, sleeveless, pink frilled gloves, black belt, purple eyes, purple hair, pink thigh boots

🎲 Optional prompts

star hairclips, high heel boots, multicolored eyes

🐣 For beginners

If you like my posts, click on the image to view details. All the images I share are original ComfyUI-generated images with no Photoshop post-editing. Some works with numerous highly restored fine details use inpainting. If you don’t have extremely strict requirements for image quality, regular Hi-res upscaling will deliver great results. I usually set the LoRA weight between 0.8 and 1. For some highly fitted LoRAs, lower the weight to 0.6–0.7 when changing outfits. To replicate the style of my posts, please use my recommended checkpoints.

Feel free to try a variety of checkpoints within the Illustrious and NoobAI series—most have great compatibility. I’ve tested quite a few of them, though I rarely use many fine-tuned checkpoints for my posts since their inherent quality is just average. If you have solid prompt skills, you can easily master many kinds of art styles, not just the ones I showcase here.

💡 Tips

✦ Using Hi-res Fix can achieve better detail performance.

✦ Load the LoRA and inpaint with the trigger word for perfect results. INPAINT is very very important!

✦ If the result is very different from the intended character, the LoRA may not have loaded properly or its weight is incorrect.

✦ If the detail quality is unsuitable, try increasing the weight (without overfitting), add more precise descriptions as needed, or include negative prompts for elements you want to remove.

✦ Different base models may render details differently, and inherent model biases are difficult to avoid. Adjust your prompts if the generations are not ideal.

✦ All LoRAs trained on the Illustrious base model are theoretically compatible with NoobAI.

✦ Lowering the weight changes a lot, you know.

D o   w h a t e v e r   y o u   w a n t ！ 🚀✨🌈