Despite voice being a natural human interface, it remains underutilized in AI applications today. OpenAI aims to change this with its latest advancements, empowering businesses and developers to build sophisticated voice agents. These systems can operate autonomously, assisting users in diverse scenarios such as customer support, language learning, and accessibility tools.
What’s New?
🔹 Speech-to-Text Models: The GPT-4o Transcribe and GPT-4o Mini Transcribe models outperform OpenAI’s previous Whisper models, offering significant improvements in transcription accuracy and efficiency.
🔹 Text-to-Speech Model: This innovation allows precise control over not just the spoken words but how they are said, enhancing the expressiveness of AI-generated speech.
🔹 Agents SDK Enhancements: Developers can now seamlessly convert text-based agents into voice-driven systems, enabling natural and fluid interactions.
🔎 OpenAI highlights two approaches to building voice AI:
Speech-to-Speech (S2S): Maintains nuances like intonation and emotion, ensuring natural interactions.
Speech-to-Text-to-Speech (S2T2S): Easier to implement but may lose key details and add latency.
With affordability and accessibility at the forefront, OpenAI’s new models are poised to drive widespread adoption. The GPT-4o Transcribe model is priced at $0.006 per minute, while the GPT-4o Mini Transcribe is available at $0.03 per minute. These updates underscore OpenAI’s commitment to making voice a key focus area for AI development.
Get in touch
