Summary
Neil Zegidua argues that voice is moving from app-triggering commands toward an operating-system-like interface as software interaction shifts from machine language to human language. He traces the path from Siri-style constrained systems to LLM voice agents, then identifies low latency, paralinguistic understanding, overlapping speech, and noisy real-world environments as the next technical challenges.
Key Topics
5 key topics from Neil Zeghidour at HumanX Amsterdam 2026. Thicker branches were mentioned more often in the talk.
Voice is the new OS
Click a topic to open it
Key Takeaways
- Design voice interfaces around agents, not simple commands.
- Reduce latency by questioning cascaded speech-to-text-to-speech pipelines.
- Preserve paralinguistic cues instead of flattening speech into text.
- Support interruptions, backchanneling, and overlapping conversational dynamics.
- Prepare voice AI for noisy, multi-speaker real-world settings.
“voice becomes the most natural way of communication.”
N
Neil Zeghidour
Speaker
Share
Up Next
RRM
WatchNext in agenda
Engineering the Physical World
Rebecca Schaevitz, Rob Pegoraro, Maria Dunford, Fabrizio Pirondini
physical aidigital twins
SRA
WatchAlso on voice ai
Emotion, Intent and the Future of Voice AI
Shawn Wen, Renate Nyborg, Alex Gay, Ivan Mehta
voice aiconversational ai
AS
WatchAlso on robotics
The Race to Build Real-World Intelligence
Alex Kendall, Sharon Goldman
autonomous drivingphysical ai