# OpenAI Launches Customizable Text-to-Speech Models for Developers

OpenAI announced a major upgrade to its audio API, introducing next-generation text-to-speech models with an unprecedented feature: developers can now instruct the AI on exactly how to speak.

For the first time, the API allows natural language instructions to control vocal delivery. Developers can simply tell the model to "talk like a sympathetic customer service agent" or presumably any other speaking style, rather than being limited to preset voice options.

This breakthrough represents a significant leap in voice AI customization. Previously, text-to-speech systems typically offered only basic controls like voice selection, speed, and pitch. The new capability enables developers to fine-tune emotional tone, speaking style, and personality through simple text prompts.

The implications are substantial for voice-enabled applications. Customer service chatbots could adapt their tone based on context—sounding reassuring during complaints or enthusi