GPT-4o: natively multimodal
One model handles text, images and voice, with real-time spoken conversation, interruptions and emotion.
~320 msaverage voice response time, close to a human
Why it matters
Interaction shifted from typing to talking, and GPT-4-level capability became free.
Costs and controversy
A demo voice that resembled an actress drew controversy and was withdrawn.
Built on
Led to
Source: openai.com ↗