Perception · Iris-2
Iris.It sees youseeing it.
Iris is the perception model. It reads expression, gaze, posture and what's on a shared screen, and turns it into context the conversation can use, in real time.
- Inputs
- Camera, screen share, audio prosody
- Signals
- Emotion, engagement, gaze, environment
- Latency
- Per-frame, streamed to the LLM layer
- Queries
- Custom visual & screen awareness questions
Inside the model
What it doesfor a conversation.
- Emotion, not just words
- A pause with a frown means something different from a pause with a smile. Iris passes those cues to the brain so replies can soften, slow down or check in.
- Screen awareness
- When a user shares a screen, Iris understands what's on it: an error dialog, a spreadsheet, a slide. Support and tutoring personas can talk about what they see.
- Your own questions
- Ask Iris anything visual in plain language: 'Is the candidate reading from notes?' 'Does the patient look uncomfortable?' Answers arrive as structured events.
Other models
ManavOS · Iris