OpenAI Preparing GPT-Bidi-1 for Next-Generation ChatGPT Voice Upgrade

OpenAI is preparing GPT-Bidi-1 (tentative name), a next-generation audio model, for a major upgrade to ChatGPT's voice mode. This model is designed based on a bidirectional (Bidirectional) architecture, enabling it to listen and speak at the same time, recognize interruptions during conversation, and adjust its responses mid-sentence, allowing for much more natural conversations.

This update aims to close the gap in voice technology, which has stagnated compared to the pace of advancement in text models. OpenAI views voice as likely to become the primary means of communication with AI over text, and is focusing on developing audio-centric hardware and supporting tools toward this goal.

Key features and expected changes are as follows:

• Intelligence Level Selection: As with existing text models, users will be able to choose between High, Medium, and Instant modes depending on the task, adjusting the balance between speed and depth of reasoning.
• Mode Switching: Users are expected to be able to switch between the existing Advanced Voice Mode and the new Bidi mode.
• UI Improvements: The recent change allowing the voice bubble to be dragged to the center of the screen is interpreted as part of this redesign.