Skip to main content
ExploreExplore
All use cases

SOLUTIONS / AI companions

AI companions

Combine a character profile, conversation history, and the current message to keep exchanges coherent.

Explore
Built forAI companion teamsRole-play platformsInteractive entertainment teams

01Multi-turn dialogue

Every turn builds on the last.

Using the history your app provides, the model follows references, emotions, and topics across the conversation.

Input
Character profile, conversation history, current message
Output
A reply for this turn that follows from earlier context and fits the character
View workflow
Multi-turn dialogueBased on the history your app provides

I’m giving my first public talk tomorrow, and the opening worries me most.

Character

The opening can be the hardest part. Are you more worried about forgetting your words or getting started?

Earlier context in this request

The current message gives the time of the talk, the task and the worry.

02Role-play

New story. Same character.

A character’s motives, voice, and boundaries stay consistent as the story unfolds.

Input
Character profile, story background, current event
Output
Dialogue and intended actions consistent with the character’s motives
View workflow
Character conversation artwork
Character profile01 / PERSONA

The traveling friend

Curious and candid, with a knack for turning an ordinary day into a good story.

“I haven’t planned the route yet. But I’d like to choose the first stop with you.”

03Audiovisual conversation

Expression beyond words.

Expressions and movement bring listening and responses to life. Your app and the dialogue model manage the conversation.

Input
Speech rhythm, character reference, movement requirements
Output
Continuous character expression and conversation state
View workflow

IN PRACTICE

Pick up the conversation

In the same conversation, a user discusses a choice made by a character in a novel and later adds a new reading of it. The app sends the earlier dialogue along with the new message; the model draws on what was said before to ask why the user’s view changed, while the character video shows listening and thinking.

What to prepare

  • Character personality, speech habits and background
  • The dialogue history and current message sent with this request
  • The current message, voice and preferred conversational style

Deliverables and feedback

  • Dialogue responses true to the character’s personality
  • Follow-up responses consistent with earlier facts and the character’s stance
  • Facial expression, posture and movement matched to the tone

CONNECTED INTELLIGENCE

Models working together.

Connect your business context, model outputs, and application feedback.

Understand people and context

OmniSoul

Generates context-aware responses from the character profile, conversation history, and current message supplied by your app.

Input
Character profiles, conversation history, and context
Output
Context-aware responses and proposed actions

Animate characters in real time

OmniFlow

Turns the rhythm of speech and the motion directions in a response into a coherent character performance.

Input
Reference images, speech, and motion prompts
Output
Character video with coordinated speech and movement

  1. Set the character and context

    Your app supplies the character profile, relevant conversation history, and current message for each request.

  2. Generate a context-aware response

    OmniSoul combines the dialogue history in the request with the current topic to produce a spoken response and action intents.

  3. Bring the response to life

    Pass speech and motion prompts to OmniFlow. Your app manages the conversation record and includes relevant history in later requests.

FROM PILOT TO PRACTICE

Test what works.

Test representative tasks at a small scale. Record inputs, model versions, and results, then refine the solution based on what you learn.

Access and availability depend on the model and release.

Character consistency

Check that the character setup holds up across rephrased questions, different topics and many turns.

Contextual consistency

Check that responses correctly understand the supplied context, resolve references and update their judgment as the user adds information.

Natural interaction

Observe how listening, pauses, emotional responses and character movement fit the current context.

Start a pilot

Tell us about your users, tasks, and success criteria. We’ll help you explore the right models.

Research and industry resources

Related use cases

All use cases
Concept image for Games and interactive stories

Games and interactive stories

Characters respond to player choices, world events, and changing relationships.

Concept image for Learning and training

Learning and training

Practice real conversations with clear goals and actionable feedback.

Concept image for Brand assistants and guides

Brand assistants and guides

Answer questions and guide visitors with verified product and service information.

OMNISOUL · RESEARCH PREVIEW

Request access

Tell us about your team and your research.

Research area