AudioPaLM: language-backbone initialization for unified speech/text generation

What changed

Combined a text-language backbone with audio-token modeling to transfer text pretraining into speech understanding, generation and translation.

Scientific scope

Combined a text-language backbone with audio-token modeling to transfer text pretraining into speech understanding, generation and translation. Explicitly inherits PaLM2, AudioLM and neural codecs. Zero-shot source/target combinations differ from no training for a language or zero speech supervision. No voice clone or unqualified current model superiority inferred.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Explicitly inherits PaLM2, AudioLM and neural codecs. Zero-shot source/target combinations differ from no training for a language or zero speech supervision. No voice clone or unqualified current model superiority inferred.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• AudioPaLM: A Large Language Model That Can Speak and Listen

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction