Hello! This LoRA was designed to generate high-quality hentai images of the character Rio-chan (based on the work Dearest Blue – Ootori Rio); I’ve refined it since the last version (v6) with more comprehensive training. Now, the outfits she wears in the hentai are much more consistent! I’ve listed the tags I use for each outfit below!
For the best results, I strongly recommend using the waiIllustriousSDXL or REED_XXX_illustrious_SDXL checkpoints—in my opinion, the two best models for generating anime-style images!
I always use a standard prompt structure to start my work (*POSITIVE PROMPT*), and I think you could use it as a base for your own! POSITIVE PROMPT AND KEYWORDS – "Masterpiece, superior quality, ultra-detailed, impeccable aesthetics, sharp focus, anime screenshot, 2023, anime coloring,
1 girl: <lora:riochan-v13-IL:0.9> red hair, purple eyes, long hair, medium breasts, slight blush..."
1. If you prefer, you can replace medium breasts with large breasts.
2. For more creative freedom, set the LoRA weight between 0.8 and 0.9; for a look closer to hentai, use 1 or 0.95.
3. I strongly recommend using ADetailer to improve the visual quality of the eyes (running it with default settings yields great results).
4. In most cases, I use these configuration settings:
Sampler Method: Euler a
Steps: 20 to 25
Hires. fix: R-ESRGAN 4x+ Anime6B
CFG - 5 to 7
PROMPTS FOR COSTUMES
IDOL - 1 girl, idol outfit, sexy idol, idol uniform, breasts, solo, cleavage, navel, skirt, (red skirt:1.25), bow, off-shoulder, midriff, hair bow, blush, collarbone, miniskirt, red bow, black hairband, black belt, red collar, gold chain on the right side of the skirt, black bra with red stripes, pink ruffle, zettai ryouiki, black thigh-high socks,
COLLEGE - college uniform, white t-shirt, red tie, black skirt with white details, long-sleeved shirt, black boots,
PAJAMAS - Pink pajamas,
Later on, I will write an article about other variations—such as manga style and cover art—using only prompts*, without the need for additional LoRAs.
This is my new version, a revision of Rio-chan version 6.0, and I believe you will get much better results than before! Since I trained her with more epochs and greater creative freedom, she fits into MANY different scenarios—both NSFW and SFW. Go ahead and make the most of it, haha.
Tag me in your posts! I want to see how you fantasize and imagine adventures with her! It really excites me...
- Use these aspect ratios for the best results!
768x1024, 1024x768, 768x768, 1024x1024, 512x1024, 1024x512, 768x1350, 1350x768
Description
This LoRA is designed to generate high-quality hentai images of the character Rio-chan (based hentai Dearest Blue - Ootori Rio), focusing on consistent and detailed character reproduction. It's perfect for creating explicit and suggestive art of her, capturing her iconic appearance in various adult scenarios. I love imagining it!
For best results, it's highly recommended to use waiIllustriousSDXL as a checkpoint, as LoRA was specifically trained in its "illustrious" style and the two work in perfect synergy.
I always use this prompt to start my work (POSITIVE PROMPT), and I think you can use it as a base for yours!
POSITIVE PROMPT AND KEYWORDS - "Masterpiece, superior quality, ultra-detailed, impeccable aesthetics, sharp focus, anime screenshot, 2023, anime coloring, 1 girl: <*LORA RIO*> red hair, purple eyes, long hair, teen, very blush, (additional)..."
*Later I will write an article for other variations such as manga and covers using only prompts without the need for additional loras.
This is my new version, a 6.0 revision of Rio-chan, and I think you'll get much better results than before! Because I trained it with more eras and greater creative freedom, it fits into MANY different NSFW and SFW scenarios. Take advantage and explore this to the fullest, haha.
Tag me in your posts! I want to know how you fantasize and imagine adventures with her! That excites me...
NOTE - Regarding more technical things - I think this can help achieve good results.
Checkpoint - waiIllustriousSDXL
- Sampling steps - 20 to 25, no more than that is needed in my tests
- Sampling method - DPM++ 2M karras
- CFG - 5 to 7 (for me the ideal and what I always use is 5.5)
- Use these proportions for your work to obtain better results! 768x1024, 1024x768, 768x768, 1024x1024, 512x1024, 1024x512, 768x1350, 1350x768
To use hires.fix to increase the resolution, download this model: R-ESRGAN 4x+ Anime6B (from experience and based on the dimensions I use, hires.fix 1.5 to 2 produce the best results most of the time I use 1.5 or 1.6)
FAQ
Comments (6)
Is lora not set to generate ?
@foponi3097918 Yes, it is! Couldn't generate it after downloading? Tell me what's happening and I'll try to help.
@Rio_chan I see it now
Are you planning to, or will you in the future do a Lora of Yoshikawa Noa from the same series?
Hiiii, how are you? Sorry for the delay in replying! I'm learning to work with AI (flux, SD, SDXL, Loras) as a hobby, but I can add it to my list if you want <3 to see a version of Noa Chan! I think, with what I've already done, it won't be as difficult as the first Loras I created, haha
How would describe eye color ? I tried purple like in your example, but I get a different result.