Descript Overdub 2.0: AI Voice Editing Goes Professional
Professional podcasters and video creators can now edit their voice recordings like text documents, with AI seamlessly generating missing words in their own voice.
Revolutionary Audio Workflow
Descript's Overdub 2.0 lets creators edit voice recordings like text documents. Fix a mistake, insert a clarification, or update a dated reference, and the AI generates the missing words in the original speaker's voice — usually convincingly enough to pass a careful listen.
Production time for podcast and video work drops sharply, which matters most to independent creators who do their own editing.
The Creator Economics
Re-records used to be the tax paid for errors made during recording. Eliminating that tax changes the economics of episodic content: more episodes, shorter turnaround, and a lower bar to entry for creators who used to find the edit cycle exhausting.
The enterprise side — corporate training, help content, internal comms — moves faster still, because consistency across a large library becomes tractable.
The Consent Conversation
Overdub requires the speaker's voice to be trained, and Descript has thoughtful consent flows. Broader ecosystem concerns remain: what happens when a publisher wants to keep editing an archive after a contributor leaves, or when a voice model outlives its owner's wishes.
These are policy questions, not engineering ones. They will be worked out contract by contract, and the defaults will matter.
