`OpenAI API` adds new realtime voice models for reasoning, translation, and transcription

Voice input now spans reasoning, translation, and transcription in one realtime API flow. This is immediately useful for shipping tighter voice UX, though pricing and latency still decide adoption.

• —The new voice models handle reasoning, translation, and transcription on spoken input, reducing multi-step pipeline glue.
• —Realtime support points to more natural turn-taking and lower-friction voice interfaces for assistants, support, and language tools.
• —The practical win is API-level consolidation: fewer separate speech components to orchestrate inside OpenAI API voice stacks.