AI voice interaction is undergoing a "dimension-reducing" evolution. Recently, many users discovered a new voice model called "Bidi1" on the web and app versions of ChatGPT, indicating that OpenAI is preparing for a larger-scale test, bringing an unprecedented smooth experience to AI voice interaction.
For a long time, AI voice assistants have followed a "I ask, you answer" linear logic, where users had to wait for the AI to finish outputting the previous message before initiating the next interaction. The emergence of the Bidi1 voice model has completely broken this constraint. Its core highlight is "bidirectional parallel processing": the AI can not only listen to user input in real-time while speaking, but also immediately respond to user interruptions or new instructions during the conversation.

This interaction mode greatly narrows the gap between human-computer dialogue and real human communication. In a demonstration case, when the model was performing the task of "counting from 1 to 10," the user could interrupt at any time to ask it to "count down," and the model could seamlessly switch to the new instruction. This "listening and responding simultaneously, real-time response" interaction logic has completely eliminated the rigid waiting period, making the conversation extremely natural and smooth.
In terms of interface operation, Bidi1 has a high level of distinguishability. When users select this option in the model selector settings, the original voice bubbles will turn into eye-catching yellow, indicating that the user has switched to this advanced voice mode.
Although OpenAI has not officially released this feature on a large scale, based on current test feedback, the launch of this function is approaching. This round of upgrades in ChatGPT not only improves the efficiency of voice interaction but also takes an important step forward in the immersion of human-computer collaboration. For users who are accustomed to handling tasks through voice, a smarter assistant that understands and responds quickly is about to be within reach.




