Zhipu AI end-to-end full-duplex realtime speech suite featuring ultra-low latency chatting and expressive voice styles.
Free full-duplex realtime voice conversation and multi-speaker reading in Web and Mobile apps
ChatGLM Voice (GLM-4-Voice) is Zhipu AI's native end-to-end speech multimodal model built directly on the GLM architecture. Departing from fragmented pipelines that daisy-chain ASR, text LLM, and TTS, GLM-4-Voice processes speech directly in the continuous audio token space. Users can engage in low-latency full-duplex calls, interrupting the AI naturally at any moment. Offering multiple expressive timbres—from professional anchors to empathetic companions—it dynamically senses user emotional cadences for seamless voice interaction.