Fact · Voice and AI
The Moshi paper says current spoken dialogue systems rely on pipelines of independent components, and it casts spoken dialogue as speech-to-speech generation instead.
Checked against its source on October 17, 2026
In the source's own words
“Current systems for spoken dialogue rely on pipelines of independent components, namely voice activity detection, speech recognition, textual dialogue and text-to-speech. ... Moshi solves these independent issues altogether by casting spoken dialogue as speech-to-speech generation.”
Moshi: a speech-text foundation model for real-time dialogue, Alexandre Défossez and and colleagues at Kyutai, arXiv (Kyutai), 2024, Abstract.
Open the original- What it measures
- Kyutai's 2024 paper introduces Moshi, a model that generates speech tokens directly from its input speech, with a text language model as its backbone.
- What it does not show
- The paper is a preprint and the comparison describes the typical design it argues against, not every commercial product. Speech-to-speech here means one model handling audio in and audio out.