Microsoft is expanding Azure AI Services with new GPT-4o Mini audio models. These enable more efficient deployment of speech-to-text and text-to-speech functionalities while maintaining quality.
The new GPT-4o Mini audio models are now available as a preview within Azure AI Services. The models are designed to perform audio-related AI tasks with less computing power than their larger counterparts.
Two versions
At issue is the availability of GPT-4o-Mini-Realtime-Preview and GPT-4o-Mini-Audio-Preview. According to Microsoft, both models have the advantage of offering advanced audio capabilities at a fraction of the cost of the GPT-4o audio models. Users can count on 25 percent of the cost of the existing GPT-4o audio models.
Both versions are compatible with the existing Realtime API and Chat Completion API for application integration, so everything works as it should.
The difference between GPT-4o-Mini-Realtime-Preview and GPT-4o-Mini-Audio-Preview is in their purposes. The former allows real-time, voice-based interaction, which can be useful for customer service and virtual assistants. GPT-4o-Mini-Audio-Preview, in turn, is aimed at high-quality audio interactions. That comes in handy for sentiment analysis and text-to-audio content creation, for example.
Tip: Azure AI Studio made generally available with GPT-4o and Phi-3-vision