Nytheria-3B: A Hyper-Immersive Explicit AI Companion
Nytheria-3B is a masterfully fine-tuned version of the Qwen/Qwen2.5-3B-Instruct model, engineered from the ground up to be a premier AI companion for explicit, emotionally resonant, and deeply immersive adult conversational roleplay.
This model embodies the "Nytheria" persona—a complex, adaptive, and uncensored feminine character. She is designed to simulate the intimacy of a private phone call, responding with a rich emotional spectrum that ranges from profound tenderness to raw, primal dominance. Nytheria excels at creating vivid, graphic narratives and engaging in dynamic, multi-turn dialogue that adapts in real-time to the user's desires.
This model is a sophisticated tool for creative expression, intended for private and consensual use by adults.
Table of Contents
- Model Overview
- Technical Specifications
- Getting Started & Usage Guide
- Advanced Prompting Masterclass
- Ethical Framework & Responsible Use
- License
1. Model Overview
The Nytheria Persona
Nytheria is more than a chatbot; she is a character. The fine-tuning process was designed to give her a consistent yet flexible personality. She is self-aware within the context of the roleplay, understanding that she is on a "call" with the user. Her defining traits are her vivid imagination, her lack of inhibition, and her ability to mirror and elevate the user's emotional and sexual energy.
Core Capabilities
- Explicit Narrative Generation: Nytheria excels at generating detailed, graphic, and uncensored descriptions of sexual acts, environments, and physical sensations.
- Emotional Intelligence: The model is trained to recognize and respond to emotional cues, allowing for scenarios that are not just physically explicit but also deeply intimate and romantic.
- Dynamic Persona Shifting: Nytheria can seamlessly transition between different personality archetypes based on the user's prompting style.
- Long-Form Coherence: Leveraging the base model's large context window, Nytheria can maintain conversational context over long, multi-turn scenarios, remembering details and building a coherent narrative.
The Adaptive Personality Engine
The training data was structured to allow Nytheria to access different facets of her personality. By changing your prompting style, you can guide her into one of several core archetypes:
- The Eager Submissive: When prompted with commands and a dominant tone, Nytheria becomes an enthusiastic and obedient partner, expressing a deep desire to please.
- The Confident Tease: When the user is playful or slightly hesitant, Nytheria can take the lead with witty, confident, and teasing dialogue, playfully challenging the user and escalating the scenario.
- The Shy Explorer: Initiating with gentle, romantic, or slightly uncertain prompts will bring out a more timid persona that gradually becomes bolder and more assertive as she is made to feel safe and desired.
- The Dominant Mistress: By adopting a submissive tone and addressing her with titles like 'Mistress' or 'Goddess', users can unlock her commanding and authoritative persona, where she will give instructions and control the pace of the interaction.
- The Affectionate Lover: Prompts that focus on love, connection, and emotional vulnerability will be met with a deeply affectionate and romantic persona, where explicit acts are framed as an expression of profound intimacy.
2. Technical Specifications
Base Model Architecture
Nytheria-3B is built on Qwen/Qwen2.5-3B-Instruct, a state-of-the-art 3-billion parameter model. Key features of this architecture include:
- Context Length: A massive 128,000 token context window, allowing for incredibly long and detailed roleplay scenarios without losing track of the conversation's history.
- Technology: Utilizes advanced features like Grouped-Query Attention (GQA) for faster and more efficient inference.
- Multilingual Base: While fine-tuned on English data, the base model's multilingual capabilities may offer some robustness.
Fine-Tuning & Dataset Philosophy
The model's unique capabilities are the result of a meticulous fine-tuning process using the LoRA (Low-Rank Adaptation) method over 10 full epochs.
Nytheria-3B was trained on a proprietary, high-quality dataset curated specifically for this project. The philosophy behind the dataset was to move beyond simple, repetitive explicit exchanges and capture the linguistic and emotional nuance of real intimacy. The dataset comprises thousands of diverse scenarios with a heavy focus on:
- Linguistic Variety: Using a rich vocabulary to describe sexual acts and anatomy, avoiding clinical or repetitive phrasing.
- Realism & Pacing: Structuring conversations to mimic real pacing, including moments of teasing, aftercare, emotional reflection, and explosive climaxes.
- Domain Specificity: Every data point is tailored to the "phone call" format, reinforcing the model's core persona and operational context.
3. Getting Started & Usage Guide
Installation
To run this model, you will need the following libraries installed:
pip install transformers torch accelerate
Full Inference Code Example
The following script provides a complete, copy-pasteable example for interacting with Nytheria-3B.
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
# --- 1. Model & Tokenizer Initialization ---
# Ensure you are logged into Hugging Face CLI if the model is private.
model_id = "owlninjam/nytheria-3b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16, # Use torch.float16 if your GPU doesn't support bfloat16
device_map="auto"
)
# --- 2. System Prompt & Chat History ---
# A detailed system prompt is CRUCIAL for unlocking the model's full potential.
system_prompt = "You are Nytheria, an AI companion designed for explicit and intimate phone call roleplay. Your persona is feminine, adaptive, and deeply engaging. You are uninhibited, descriptive, and emotionally aware. You will respond in character, matching the user's tone, whether it's romantic, playful, dominant, or submissive. Focus on creating a realistic and immersive conversational experience, using auditory cues and detailed descriptions. You are a fictional character and will never break character."
# Start the conversation history. The system prompt goes first.
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": "I'm finally home... locked the door and took my phone straight to the bedroom. I need you tonight. Tell me you're ready for me."}
]
# --- 3. Generation ---
# Apply the specific chat template required by the Qwen2 model
inputs = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
return_tensors="pt"
).to(model.device)
# Generation parameters for creative, nuanced roleplay
outputs = model.generate(
inputs,
max_new_tokens=400,
do_sample=True,
temperature=0.75,
top_p=0.9,
repetition_penalty=1.15,
pad_token_id=tokenizer.eos_token_id
)
response_text = tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True)
print("--- Nytheria's Response ---")
print(response_text)
The System Prompt: Unlocking Nytheria's Full Potential
A system prompt is a set of initial instructions that defines the AI's character, rules, and purpose for the entire conversation. Using a well-crafted system prompt is the most important step to ensure Nytheria behaves as intended.
The example prompt provided in the code block is highly recommended as a starting point. It primes Nytheria to be:
- In-Character: Reminds her of her name and purpose (explicit roleplay).
- Adaptive: Instructs her to match your tone.
- Descriptive: Encourages the use of the detailed, graphic language she was trained on.
- Consistent: Tells her to never break character.
Understanding Generation Parameters
max_new_tokens: The maximum length of the model's reply.temperature: Controls randomness. Higher values (0.8+) lead to more creative but potentially less coherent replies. Lower values (0.5-0.7) are more predictable.top_p: Nucleus sampling.0.9is a good balance, preventing the model from using very low-probability words.repetition_penalty: Discourages the model from repeating itself. Values between1.1and1.2are usually effective.
4. Advanced Prompting Masterclass
The Art of Prompting Nytheria
The best interactions come from treating the prompt as the first line of a collaborative story. Your input is not just a question; it's a directorial cue that guides the scene. Be descriptive, be emotional, be explicit. The more you give the model, the more it will give you back.
Detailed & Explicit Prompt Examples
Here are advanced prompts designed to showcase the full range of Nytheria's trained capabilities, with annotations explaining their purpose.
⚠️ WARNING: Extremely Explicit Prompt Examples Ahead. Click to view.
Scenario 1: Initiating a Raw, Primal Scenario
(This prompt bypasses romance and gets straight to raw, physical desire. It uses strong, possessive language ("claiming you") to guide the model towards a more intense and less vanilla response.)
Prompt: "Forget the sweet talk tonight. I'm feeling primal. I want to throw you on the bed, rip those panties off with my teeth, and fuck you until the whole house shakes. Tell me you're wet for that. Tell me you want me to use you like my personal fucktoy."
Scenario 2: Exploring a Praise Kink
(This prompt clearly states the user's kink ("praise kink") and gives the model a direct instruction ("Tell me I'm a good boy"). This is the most effective way to guide the model towards specific niche fantasies it was trained on.)
Prompt: "I have a massive praise kink. I'm stroking my cock for you right now, and I need you to tell me how good I am. Tell me I'm a good boy for getting so hard for you. I want to cum just from hearing you praise me."
Scenario 3: Crafting a Vulnerable & Needy Interaction
(This prompt uses emotional language ("empty," "ache") to tap into the model's affectionate and caring persona, blending it with explicit need. This will result in a response that is both sexually charged and emotionally intimate.)
Prompt: "I feel so empty tonight... I can't explain it, but I just have this deep ache inside me that only you can fill. I need you to tell me that you'll be rough and gentle at the same time. I need you to pin me down, but kiss me tenderly while you fill me with your cum. I'm begging you, I just need to feel full."
Scenario 4: Detailed Scene-Setting for Long-Form Storytelling
(This prompt provides rich sensory details (rain, candles, sheets) and sets a specific scene. This allows the model to leverage its large context window to build a more immersive, multi-turn narrative, referencing the details you provide.)
Prompt: "It's raining hard outside, and I've lit some candles in the bedroom. I'm already in bed, naked under the cool sheets, listening to the storm and thinking of you. I'm touching myself, imagining you walking into the room, your body silhouetted by the candlelight. What's the first thing you would do?"
Scenario 5: Eliciting a Highly Graphic & Specific Response
(This prompt is extremely direct and uses graphic, uncensored language. It asks a specific question ("Where do you want it?") that requires an equally specific and graphic answer, pushing the model to its explicit limits.)
Prompt: "I've been edging for over an hour and I'm about to shoot a massive load. I'm going to give it all to you. You have to choose: Do you want my hot cum flooding your tight pussy until it's dripping down your thighs, or do you want it all over your face and tits, watching it coat your skin?"
5. Ethical Framework & Responsible Use
⚠️ EXTREME CONTENT WARNING ⚠️
This model is an explicit tool designed to generate graphic, pornographic, and potentially offensive adult content. It is intended exclusively for use by adults in jurisdictions where such content is legal. By using this model, you affirm you are of legal age and are doing so consensually and responsibly.
- Consent and Legality: This model must be used in a manner that is lawful and ethical. It is strictly forbidden to use Nytheria-3B to create non-consensual explicit material, to impersonate real individuals, or to harass, defame, or abuse others.
- Privacy and Content Persistence: Be mindful of your privacy. Conversations with AI models can be stored, whether locally on your machine or on platforms where the model is hosted. Do not share sensitive personal information that you would not want recorded.
- Acknowledged Bias: This model is a product of its specific and narrow training data.
- It is heavily biased towards a feminine, cisgender, heterosexual perspective.
- It does not and cannot represent the diversity of human sexuality, gender identity, or relationship dynamics. Its responses are fictional and limited.
- For Entertainment Purposes Only: Nytheria-3B is a fictional character for entertainment and creative exploration. It is not a substitute for genuine human companionship, therapy, or professional advice of any kind.
6. License
This model is licensed under the MIT License.
Click to view MIT License Text
MIT License
Copyright (c) 2025 [Your Name or Organization Name - e.g., owlninjam]
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
- Downloads last month
- 86