SteerDuplex
AI model · Multimodal model · Text → Text by Scale AI
Steerable Duplex Speech Dialogue Models for Natural AI Interaction
A steerable duplex speech dialogue model from Scale Labs designed for natural, multi-turn AI voice interactions with simultaneous listening and speaking capabilities.
Model specs
The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.
- Languages
- English
Category: Conversational AI
- Modalities
- Audio (Speech) → Audio (Speech)
- Released
- 2026-09-17
- Model Category
- Multimodal
- Release Date
- September 17, 2026
- Interaction Type
- Duplex Speech Dialogue
- Developer
- Scale Labs
- Steerable duplex speech dialogue model
- Enables multi-turn spoken interactions
- Developed by Scale Labs research team
- Released September 17, 2026
Overview
SteerDuplex is a steerable duplex speech dialogue model developed by Scale Labs, the research arm of Scale AI. Released on September 17, 2026, the model is designed to improve the naturalness of AI voice interactions by enabling multi-turn, simultaneous dialogue capabilities. It focuses on steerability, allowing developers to control the flow and style of speech interactions for more realistic verbal exchanges. SteerDuplex represents Scale's research into multimodal speech systems that can handle complex, real-time conversations. The model is part of a broader research portfolio at Scale Labs that includes agents, post-training, and safety alignment.
Speech Interaction
- Steerable Duplex Speech: The model architecture supports steerable duplex speech dialogue for natural, multi-turn verbal exchanges.
- Multi-turn Interaction: SteerDuplex is designed to handle multi-turn interactions in spoken dialogue systems.
- Duplex Capability: The model enables simultaneous listening and speaking capabilities for more realistic AI voice interactions.
Research & Development
- Multimodal Speech Architecture: SteerDuplex is a multimodal model developed by Scale Labs focusing on speech dialogue.
- Steerable Control: The research focuses on steerability, allowing for controlled flow and style in speech interactions.
Use cases
- Developers use SteerDuplex to build voice assistants that can handle simultaneous speaking and listening for more natural flow.
- Research teams utilize the model to study steerability in multi-turn spoken dialogue systems.
- Enterprises implement the model to create more realistic and responsive AI-driven customer voice interactions.
Specifications
- Modalities
- Audio (Speech) → Audio (Speech)
- Released
- 2026-09-17
- Model Category
- Multimodal
- Release Date
- September 17, 2026
- Interaction Type
- Duplex Speech Dialogue
- Developer
- Scale Labs
Added by AntSpace from public sources (labs.scale.com).