The news from TechCrunch was caught and relayed by a Russian robot summix_bot

It’s been a busy week for OpenAI, with lots of news about the company. But OpenAI does not stop and already at DevDay 2024 wants to attract even more developers to create tools based on AI models.

OpenAI introduced Realtime API – a beta version for building low-latency applications with AI-generated voice responses. It’s not quite like ChatGPT’s Advanced Voice Mode, but it’s close.

In a press briefing, OpenAI chief product officer Kevin Weil said the recent departures of CTO Mira Murati and principal investigator Bob McGrew will not affect the company’s development. He noted that these people have been excellent leaders and helped OpenAI achieve today’s results and added that the company is not going to slow down.

OpenAI still strives to be the best platform for developing AI applications. They already have over 3 million developers. But the competition is growing.

OpenAI has reduced the cost of API access by 99% over the past two years. This may have been due to price cuts by Meta* and Google.

One of the new features of OpenAI is the Realtime API. It allows developers to create applications with near real-time interaction and choose from six voices from OpenAI. These voices are different from those used in ChatGPT. Developers cannot use third-party voices to avoid copyright issues.

At the demo, OpenAI’s head of developer work, Romain Hugh, showed an application for planning a trip to London using the Realtime API. It allowed users to receive responses with minimal delay. The API also has access to a number of tools that allow the program to annotate a map with the location of restaurants.

Hugh also demonstrated how the Realtime API can talk to a person on the phone to find out about a food order for an event. However, the OpenAI API cannot directly call restaurants or shops, but only integrate with services such as Twilio.

OpenAI also introduced a vision fine-tuning feature in their API. This should help developers improve the performance of GPT-4o in tasks related to visual perception.

Telegram channel of the author of the blog.

* Meta Platforms is recognized in Russia as a company that carries out extremist activities. Meta and its social networks Facebook and Instagram are banned in Russia.