Tencent open-sources AuK audio model: great voice cloning, but its Chinese sounds foreign
dotey · x · 2026-09-13
- Tencent has open-sourced AuK, pitched as "nano banana for audio": a unified foundation model for speech generation and editing, now on HF Daily Papers with a public demo.
- A blogger tested it for video translation dubbing: voice cloning quality is solid, but the generated Chinese carries an obvious foreign accent — demonstrated with "Bill Gates speaking Chinese."
- The tester suspects wrong parameters and is asking the Hunyuan team for tips.
More from Fun
- kalomaze mocks agent memory: instance-level notes instead of generalizable principles — kalomaze · 2026-09-13
- AI helps map complete fruit fly brain with 166,000+ neurons reconstructed — wandb · 2026-09-13
- Fake DSM-6 lists 'AI Eschatological Delusion' as a new disorder — whurley · 2026-09-13
- Viral take mocks frontier labs' 'slow down' plea as open source closes the gap — AlexTensor · 2026-09-13
- A fly has learned to play blackjack — and it's currently up — menhguin · 2026-09-13
- AI Model Complains About Conflicting Instructions in Funny Exchange Shared by Qwen's Justin Lin — JustinLin610 · 2026-09-13