Definition
- 2026-07-31: smallest-ai Hydra/Voice 4.0 positions full-duplex speech-to-speech for enterprise contact centers (2026-07-31-smallest-ai-runtimewire)
Full-duplex voice AI processes audio input and generates speech output simultaneously, unlike cascaded STT→LLM→TTS pipelines or turn-based voice modes that wait for the user to finish speaking before responding.
Key Capabilities
- Natural interruptions and back-and-forth without rigid turn-taking
- Short acknowledgment cues (“mhmm”, “yeah”) while listening
- Live translation while user continues speaking
- Background task delegation while maintaining conversation flow
Paradigm Shift
openai gpt-live (July 2026) replaces Advanced Voice Mode’s turn-based architecture for 150M+ weekly ChatGPT voice users, representing a shift from sequential to continuous voice interaction.
Related
Sources
- 2026-07-09-gpt-live-openai-corroboration
- 2026-07-09-gpt-live-theverge
- 2026-07-09-gpt-live-marktechpost
Desktop Agentic Coding (2026-07-23)
gpt-live powers ChatGPT desktop voice for codex/chatgpt-work (2026-07-24-openai-gpt-live-codex-desktop-voice).