OpenAI to Retire Advanced and Standard Voice Models in ChatGPT Next Month as GPT-Live Usage Spikes

Highlights

• ChatGPT will retain conversations with Advanced and Standard voice
• OpenAI’s Advanced voice is powered by a turn-based voice model
• OpenAI’s Standard voice is a speech-to-text AI model

OpenAI launched GPT-Live earlier this year, in July, as the Sam Altman-led AI giant's latest voice model. Since its introduction, the US-based company has witnessed a notable spike in its usage. Now, the ChatGPT developer has announced that it will be retiring the Advanced and Standard Voice models next month, as most users have transitioned to the new GPT-Live voice model. However, the chatbot will retain user conversations with the Advanced and Standard models. While ChatGPT Voice will no longer be powered by the two speech-to-text and turn-based voice models, select functionalities will remain based on one of the models.

OpenAI to Retire Advanced and Standard Voice Models in ChatGPT in December

In a post on X, the US-based AI giant announced that it will retire Advanced and Standard Voice models on December 11. Users will no longer be able to select either model from ChatGPT Voice's settings menu. After the transition, ChatGPT Voice will be powered only by GPT-Live. The company started notifying users about the transition on Tuesday. OpenAI says that users can continue to use the two voice models during the transition period.

The Sam Altman-led company says that since the launch of the GPT-Live model in July, the usage of ChatGPT Voice has “nearly doubled”. OpenAI claims that “virtually all” ChatGPT Voice users now use GPT-Live for conversations with the AI chatbot. Further, the tech firm says that the decision to retire Advanced and Standard Voice models will allow OpenAI to “simplify the voice experience”.

Further, OpenAI said that the change will let the company focus on “improving GPT-Live”. The AI giant said that it will bring its learnings from Advanced and Standard Voice models to GPT-Live to work on multiple fronts, such as “depth and detail of responses, maintaining continuity across conversations”, and improved personalisation. “Our ultimate goal is to make voice a seamless part of the ChatGPT experience,” the company said.

ChatGPT will retain all the conversations that users had with the two outgoing voice models, allowing users to access them in the future. To recap, OpenAI's Advanced is powered by the turn-based voice model. It is capable of natively processing and generating audio. The company highlighted that the model had to wait for the user to stop speaking to generate a response, which led to a “rigid back-and-forth”.

On the other hand, Standard Voice is a combination of three AI models, including a speech-to-text (STT) model, an LLM for generating responses, and a text-to-speech (TTS) model. OpenAI points out that the Standard Voice model offered relatively slow responses and resulted in a loss of “information across voice and text modalities”.