OpenAI Unveils Advanced Voice Agent Capabilities
2025-03-21 / Read about 0 minute
Author:小编   

On March 21st, at 1 AM, OpenAI launched three groundbreaking voice models tailored for the development of sophisticated voice AI agents during a live technical broadcast. The offerings include two speech-to-text models, namely GPT-40 Transcribe and GPT-4 Mini Transcribe, alongside a text-to-speech model, GPT-40 Mini TTS. Notably, GPT-40 Mini TTS stands out as it empowers developers to customize the emotional tone and stylistic nuances of the synthesized voice. Furthermore, OpenAI has enhanced its speech-to-text API with a robust streaming capability. This feature allows developers to feed continuous audio streams into the model in real-time, enabling instantaneous text output and responses. This real-time interaction is pivotal for applications demanding prompt feedback, such as real-time voice dialog systems and voice conference transcription.