ToolsThe story, in brief
Introducing next-generation audio models in the API
OpenAI just unlocked style-based voice customization. Developers can now tell TTS models to speak like 'a sympathetic customer service agent'—turning generic audio into persona-driven interactions.
The KeyNews take
Why it matters
OpenAI's next-gen audio APIs move text-to-speech from one-size-fits-all to instruction-based voice personas, enabling founders to ship voice agents with custom tone and personality without custom training. This is a direct competitive move in the emerging voice-agent stack.
The key facts
4 to knowText-to-speech model now accepts style instructions (e.g., 'sympathetic customer service agent')
First time developers can customize voice behavior via API prompt
Voice agent customization without retraining
Announced March 20, 2025 via OpenAI official channel
Go to the source
OpenAI Blogopenai.com
Publisher excerpt: For the first time, developers can also instruct the text-to-speech model to speak in a specific way—for example, “talk like a sympathetic customer service agent”—unlocking a new level of customization for voice agents.