GPT-Realtime-Translate is a new solution from OpenAI. The feature can translate conversations live while someone is speaking, without noticeable delay, according to the company.

 

The system supports more than seventy languages ​​as input and can translate into thirteen languages ​​as output, making it useful for customer service, among other things.

The solution helps developers build live multilingual voice communication, where every participant can speak in their preferred language and hear the conversation translated in real time, including real-time transcripts. OpenAI does not provide numerical details regarding latency, but systematically refers to it as low latency and emphasizes that the system keeps up with the speaker's pace. The company advises developers to determine the latency to measure of specific applications.

The company also presented GPT-Realtime-Whisper, a new speech-to-text feature that can transcribe conversations live as they take place. In effect, OpenAI converts audio directly into text while a conversation is still in progress.

According to OpenAI, this shifts voice AI from simple voice commands to systems that can actually think along during a conversation. The technology must not only listen, but also reason, translate, transcribe, and ultimately perform actions while a conversation is in progress.

GPT-Realtime-2, GPT-Realtime-Translate and GPT-Realtime-Whisper are available via the Realtime APIGPT-Realtime-2 costs $32 per 1 million audio input tokens ($0,40 for cached input tokens) and $64 per 1 million audio output tokens. GPT-Realtime-Translate costs $0,034 per minute. GPT-Realtime-Whisper costs $0,017 per minute.

(Open AI)

Technology