BWhat changed
Combined autoregressive semantic-token generation with a flow-matching acoustic decoder and a hybrid semantic-VQ/acoustic-FSQ codec, adapting preference training and streaming execution to this discrete/continuous architecture.
Scientific scope
Combined autoregressive semantic-token generation with a flow-matching acoustic decoder and a hybrid semantic-VQ/acoustic-FSQ codec, adapting preference training and streaming execution to this discrete/continuous architecture. Moshi, neural codecs, flow matching, FSQ and DPO retain earlier credit. Distinct from Voxtral audio understanding. Voice cloning needs consent/misuse controls; no voice generated here. Weights are non-commercial, and serving/human-preference results remain workload/source dependent.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Moshi, neural codecs, flow matching, FSQ and DPO retain earlier credit. Distinct from Voxtral audio understanding. Voice cloning needs consent/misuse controls; no voice generated here. Weights are non-commercial, and serving/human-preference results remain workload/source dependent.
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.
- Voxtral TTS
- Mistral AI: hybrid semantic-autoregressive and acoustic-flow speech synthesis — official institutional account
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