Persona guidelines
Personality · Relationships · Voice · Behavior
Tell us the character AI
you want to build.
We design the persona guidelines, prompts, and dialogue datasets your team needs to build a character AI.
System prompts
Instructions and examples for your model
Dialogue datasets
Reviewed dialogues for examples, training, and evaluation
Start a conversationExisting IP · Brand characters · New AI personas
A distinctive personality.
A responsible presence.
As AI becomes part of learning, work, and entertainment, how it treats people matters alongside what it says.
BeetlReel defines an AI character’s role, personality, relationships, and behavior around your service and its users, supporting a consistent identity and appropriate boundaries across conversations.
Services & deliverables
Ready for your development team.
Share your character concept and service goals. We produce the guidelines and data in formats agreed with your developers. Commission the full package or only the deliverables you need.
01Persona development
Interpret an existing story or brand, or develop an original character from the ground up. Connect their personality and background to choices and behavior in conversation.
Deliverables
Persona bible · Relationship definitions · Voice and behavior guidelines
02Prompt design & testing
Translate characterization into practical instructions and dialogue examples. Test and refine responses against agreed scenarios on the selected model.
Deliverables
System prompts · Example dialogues · Test findings · Implementation guide
03Dialogue dataset production
We write dialogue suited to your character and its situations. We create and review datasets for your intended use: examples in prompts, model fine-tuning, or dialogue quality evaluation.
Deliverables
Purpose-specific dialogue datasets · Data specifications · Writing and review criteria
04Dialogue evaluation & refinement
Review an existing chatbot’s voice, story consistency, and relationship cues. Identify where characterization breaks down, refine prompts and data, and evaluate again.
Deliverables
Evaluation dialogues · Error analysis · Recommended changes · Before-and-after report
Where characters meet peopleWhat experience
What experience
will you create?
Different services call for different characters and conversations.
Education & language learning
01 / 06A partner in curiosity
Practice conversation, discuss books, and work through problems. Questions, hints, and self-explanation encourage learners to think for themselves.
Games & entertainment
02 / 06Characters beyond the screen
From game NPCs to fan-facing IP characters, design responses grounded in the story world, character motivations, and relationships.
Brands & customer experience
03 / 06A voice that belongs to your brand
Give mascots and product guides a recognizable brand voice, with standards for accurate explanations and recommendations that avoid overstatement.
Exhibitions, culture & travel
04 / 06A guide who invites conversation
Tell the stories of artworks and places at a level visitors can follow, with clear distinctions between historical facts and fictional interpretation.
Professional & scenario training
05 / 06Different roles to practice with
Design simulated counterparts for interviews, negotiation, and customer service, with role-specific needs, emotions, and responses.
Everyday & digital assistance
06 / 06Assistance at the user’s pace
Guide people through everyday tasks and digital services, respecting their understanding and choices and identifying when human help is needed.
The craft behind the conversationFrom character.
From character.
To conversation.
Grounded in storytelling, writing craft, and dramaturgy, we turn a character’s motivations, values, and relationships into concrete choices and responses in dialogue.
What drives them
Define what catches a character’s attention and where they take the conversation.
What they choose
Set the principles behind agreement, refusal, conflict, and unexpected questions.
Who they speak to
Shape how closeness changes their language, including what they leave unsaid.
How they sound
Bring vocabulary, rhythm, humor, and emotional expression into actual dialogue.
Character in action“I got it wrong again. I quit.”
“I got it wrong again. I quit.”
Design what happens next.
A conversation with a child learning Hangul. As a fellow learner, Markdoong asks the child to check his reading.
General LLM
A child-friendly learning assistant
Reading 가방 together
- 👻General LLM가방이 글자를 읽어볼까?
- Child가… 밤?
- 👻General LLM가방이야. 마지막 글자 아래에 동그란 ㅇ이 있지?
- Child또 틀렸어. 나 안 할래.
- 👻General LLM틀려도 괜찮아. 어려웠던 글자 하나만 다시 연습해 볼까?
- Child하나만?
- 👻General LLM응. 방! 천천히 따라 읽어보자.
Markdoong
An alien friend learning Hangul
Reading 가방 together
- Markdoong가방이거 같이 읽어볼래?
- Child가… 밤?
- Markdoong가방! 나도 방이랑 밤이 헷갈려.
- Child또 틀렸어. 나 안 할래.
- Markdoong그럼 이번엔 내가 읽을게. 네가 맞는지 봐줘!
- Child응. 읽어봐.
- Markdoong가… 방! 맞아?
- Child맞아! 가방!
Both conversations are written illustrations of character design, not actual model outputs, records of conversations with children, or evidence of learning outcomes. A general LLM can also respond in this way.
Responsible dialogue designDesign the personality.
Design the personality.
Define its boundaries.
Turn ethical principles into dialogue rules, examples, and evaluation criteria for real situations.
Be honest
Be clear about being AI, acknowledge uncertainty, and distinguish fact from fiction. Characterization should not justify confident misinformation.
Respect choice and relationships
Warmth should not become automatic agreement or demands for an exclusive relationship. Design responses that respect user judgment and real-world relationships.
Set boundaries for the role
Define age- and context-appropriate language, personal-information requests, and responses to sensitive questions, including when to involve a teacher, caregiver, or responsible staff member.
Behavior guidelines and prompts alone cannot guarantee safety. We provide design and review criteria to support your development team’s technical safeguards and operational evaluation.
In education, make room for thinking.
Address mistakes without discouragement, offer clues before answers, and invite learners to explain in their own words. Learning outcomes depend on the audience and design; we define dialogue and evaluation criteria around learning goals rather than promise cognitive improvement.
Working together
Before we begin.
Why design dialogue when LLMs can already converse?
Fluent dialogue alone does not establish whether a character matches your intent. We define responses that fit your story and brand, then turn them into examples and evaluation criteria. When LLMs assist production, selection, editing, and review follow those standards.
Does every project need model fine-tuning?
It depends on the project. We start with prompts and examples, then evaluate the result. If fine-tuning is needed, we agree on the model and data format with your developers before preparing training data. Evaluation examples are kept separate from development and training examples. Dataset production and model training are separate tasks; responsibility for training is agreed with your development team in advance.
How are scope and fees determined?
We quote each project based on character count, source material, original writing, data volume, languages, and review rounds. BeetlReel handles design, data, and dialogue evaluation; responsibilities for integration, model training, deployment, and operations are agreed with your development team.
Let’s build a characterWhat story does
What story does
your character tell?
Extend an existing IP into AI, give a brand character a conversational voice, or create an original persona. Tell us about your source material, service goals, and the deliverables you need.
Start a conversation