BWhat changed
Released a unified speech/text interaction system with a synthetic speech-data engine, a distilled TTS model and instruction/control/tool-use training plus evaluation resources. Step-Audio2 adds a latent audio encoder and reasoning-centric reinforcement learning while integrating paralinguistic speech generation and audio/text retrieval.
Scientific scope
Released a unified speech/text interaction system with a synthetic speech-data engine, a distilled TTS model and instruction/control/tool-use training plus evaluation resources. Step-Audio2 adds a latent audio encoder and reasoning-centric reinforcement learning while integrating paralinguistic speech generation and audio/text retrieval. Inherits speech tokens, language models, distillation and tool calling. Acquired delta is the concrete integrated research system/data/evaluation resource, not first speech agent or universally production-ready behavior. The second stage inherits RL, RAG and speech-language training; it does not prove hallucination-free tool use or all-context prosody understanding. Both team fronts bundled as one system family.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits speech tokens, language models, distillation and tool calling. Acquired delta is the concrete integrated research system/data/evaluation resource, not first speech agent or universally production-ready behavior. The second stage inherits RL, RAG and speech-language training; it does not prove hallucination-free tool use or all-context prosody understanding. Both team fronts bundled as one system family.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction
- Step-Audio 2 Technical Report
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction