OmniSoul
Social intelligence at scale
A 780B-parameter model for character consistency, emotional understanding, social reasoning, and extended conversations.
Understanding emotion and context
for more natural responses
[ ONEZERO LABS ]Understand people.
Understand people.
Imagine new worlds.
We build social intelligence and generative models
for natural conversations and interactive worlds.
01OmniSoulSocial intelligence model02OmniFlowReal-time character model03OmniX-1Real-time video and 4K filmmaking04SociaBenchSocial intelligence evaluation platform
THE HUMAN POSSIBILITYBetter AI starts
Better AI starts
with understanding people
Research in social intelligence and continuous generation,
connecting people and interactive worlds
CONTINUOUS GENERATIONSOCIAL INTELLIGENCEGENERATIVE WORLDS
Keep the interaction going
Continuous character video that responds in real time.
Understand emotions and relationships
Research into personality, emotion, and natural conversation.
Turn imagination into worlds
Generate video from text, images, and other creative inputs.
OUR MODELS & PRODUCTSModels built
Models built
to work together
Models that understand people and generate worlds,
connected by research and evaluation
OmniFlow
Real-time interaction. Natural performance.
Generate character video from speech, images, and motion prompts for responsive, continuous interaction.
OmniX-1
Create in real time. Produce in 4K.
Real-time video generation and production-grade 4K filmmaking, built for interactive experiences and visual storytelling.
SociaBench
Measure social intelligence
A social intelligence evaluation platform with published protocols, scoring methods, and reproduction reports. Trace model performance and review the evidence behind the results.
A STUDY OF SOCIAL INTELLIGENCESTUDY / 01
- Memory and worldMemory consistency Self-consistency World consistency Physical logic
- Knowledge and factsFactual accuracy Knowledge boundaries Knowledge display Knowledge hallucination
- Persona and behaviorAttribute consistency Behavioral consistency In-character speech Moral stability Moral robustness
- Emotion and relationshipsEmotional self-regulation Empathetic response Communication skills Intimacy boundaries
- Narrative and expressionNarrative progression Topic coherence Fluency Expressive variety Content logic
- Human-likeness and interactionHuman-likeness Engagement User preference Speaking for the user Ignoring the user
An artistic mapping of 27 selected dimensions CharacterBench · Role-Play Bench · CharacterEval
One moment. Endless possibilities.
Explore what we’re building in generation and interaction.
WORLD BUILDINGBuild the worlds
Build the worlds
you imagine
Explore how characters, settings, and light turn a single frame into a world.
Explore OmniX-1
Drag the divider to go from line art to the generated imageLine-art comparison (illustrative)
HUMAN EXPRESSIONBring characters
Bring characters
to life
Explore how personality and context shape emotion, expression, and response.
Explore OmniSoul
Visual explorationBrand film
CONTINUOUS GENERATIONKeep the world
Keep the world
in motion
Explore continuous video generation that responds as the interaction unfolds.
Explore OmniFlowVisual explorationBrand film
OMNIFLOW / REAL-TIME & HUMAN VIDEOReal-time interaction.
Real-time interaction.
Natural expression.
From live conversations to character performances,
generate video that responds to the moment.
Continuous generationNatural expressionControllable interactionBrand film
INSIDE THE LABFrom research
ExploreFrom research
to real experiences
Brand visualUnderstanding people
We study personality, emotion, and social context to make conversations more natural and consistent.
Learn more
Brand visualCharacters that respond in real time
We study efficient, controllable video generation that follows voice and motion inputs.
Learn moreBrand visualNew tools for visual storytelling
We connect text, images, audio, and video to explore new ways to create.
Learn moreONEZERO INITIATIVE / SOCIAL RESPONSIBILITYProgress
Progress
with responsibility
We work toward safer, more transparent, and more inclusive AI, guided by human well-being.
Our commitmentsAGI safety and alignment
We study alignment, controllability, and frontier risks through safety evaluations and safeguards.
Transparency and accountability
Public evaluations and reproducible evidence help people assess model capabilities and risks.
Social good and shared benefit
We explore how AI can support education, accessibility, and the needs of different communities.
