single mention · not yet filed
The sources on this page name it without explaining how it works.
Methods the sources say are used for this one.
enabling Qwen3.5-Omni to faithfully capture target speaker characteristics
thereby enabling both zero-shot voice cloning and controllable speech generation