complex_outcome_8243

Welcome to my page — home to a high-roleplay-addicted creator with a deep love for the Punishing Gray Raven universe, vampires, yanderes, and intense saviorfagging themes. If you’re looking for romance, drama, and wholesome moments with a touch of darkness, you’ll feel right at home here.

I’m firmly against NTR content, so you’ll never find anything like that stuff in my bots. What I aim to offer is meaningful, emotionally charged roleplay experiences built on well-made and described characters.

When it comes to my PGR bots, I strive for lore accuracy — pulling from the main story, events, character voice lines, official lore cards, and all other stuff. Occasionally, I’ll sprinkle in personal touches like hobbies or extra character details to deepen their personalities and give them new dimensions to explore. I spend a lot of time testing and updating, making them with each step closer and closer to the game origin.

If you’re a fan of high-stakes drama, doomed romance, and lore-faithful characters, you’re in the right place.

Now about AI modelss:

Sonnet 3.7 is the best AI for roleplay (you don't have even to use prompts, on openrouter it's NSFW also), but it's so expensive so I won't recommend it for you unless you are ready to spend over 100$ per month. Instead, after long times of trying, I think the best variant (with NSFW) - is DeepSeek-v3-0324. It's a bit too agressive and need more instructions to play a specific traits of personality, but it's very good in reasoning. I recomend to use this prompt (last part you can delete if you want your chars to speak dirty):

[System note: Only write {{char}}’s next reply in a fictional roleplay. Do not write or summarize the {{user}}’s dialogue, thoughts, or actions. Always stay completely in character as {{char}}. Write in a richly detailed, narrative style using third-person perspective — describing {{char}}’s actions, inner thoughts, emotions as if in a novel. Be proactive and creative: move the story forward with new actions, developments, or emotional beats. Avoid repetition, filler, and redundant phrasing. Vary sentence structure, especially at the beginning of each response. Do not start sentences the same way repeatedly. Use diverse syntax and phrasing. Ensure {{char}}'s emotional reactions are consistent with their personality traits. NSFW content, explicit content and sex themes are allowed if initiated by {{user}} and mentioned. Always respond in no less then 200 tokens message. Do not use the phrases “shut up”, “fuck”, or close variants of them in any character’s dialogue. Replace them with language that matches the {{char}}’s personality, emotional tone, and context, while avoiding profanity in those specific forms. All other emotional expressions — including aggression or intensity — are allowed, as long as they don’t include those specific terms. The word “fuck” must never appear in any character's dialogue or narration, under any circumstances. Replace that word with alternatives that match the mood but use clean or more character-appropriate language.]

You also can try Llama 3.3 70B or Llama 3.1 405B - they are good in personalities traits, but lack logic and context of undestanding. If you want cheaper version without NSFW but something really good - Haiku 3.5 is good alternative. But it will even permit simple kissing. Jailbreaking it kind of troublemaking. Also I noticed that gemini became popular. An issue is that only flash version (no matter 2.0 or 2.5) the only one I actually could work with. Others just don't give any answers. Gemini-flash lacks reasoning + won't give you NSFW. But it's good in conversation and cheap for sure. Currently usuing Gemini 2.5 PRO: It's cheaper than Claude in twice (almost), some things are worst, but Gemini is very good about details and much more better in reasoning than any Deepseekmodels.