OpenAI to Retire Older ChatGPT Voice Modes on December 11
OpenAI announced that it will officially deprecate its older Standard and Advanced Voice Modes on December 11, transitioning all users to the newer full-duplex GPT-Live architecture. Historical voice chat logs will remain accessible, and users will automatically be switched to GPT-Live. This transition marks a major consolidation of OpenAI's voice interface development around full-duplex technology, reflecting a push for more natural, overlapping conversations. However, the forced migration has sparked backlash from users who prefer the depth and stability of the older models. Exceptions to the shutdown include real-time video, screen-sharing features, and custom GPT voice interactions, which will temporarily continue using Advanced Voice Mode until the December 11 deadline. OpenAI is working to improve GPT-Live's response depth and context retention following user complaints.
## BACKGROUND
Older voice modes in AI systems historically relied on a sequential pipeline of speech-to-text, a text-based large language model, and text-to-speech, requiring a turn-taking conversational style. In contrast, full-duplex architectures like GPT-Live allow the AI to process audio directly and respond while the user is still speaking.