Copilot Labs drops Scripted mode for exact voiceovers

If you have ever used AI for voiceovers, you know the headache: you feed it a crisp script, and the model starts ad-libbing, skipping words, or tweaking phrases you spent an hour polishing.

Copilot Labs just launched a fix. Mustafa Suleyman announced a new Scripted mode for audio generation, powered by Microsoft's new MAI-Voice-1 model.

Take back control of your script

The premise is delightfully straightforward. Scripted mode reads your exact input verbatim, with zero creative detours.

It sits alongside two other options. Emotive riffs slightly on your text to maximize drama, while Story performs multiple characters and voices. All three are now live inside Copilot Labs.

Why it matters

Anyone producing tutorials, ads, or scripted podcasts knows that an unpredictable voice model is useless. You end up burning credits and time just trying to get the AI to say the actual words on the screen.

We think Microsoft made the right call here by splitting generation into distinct presets. Instead of forcing one model to magically guess your intent, letting users toggle between strict fidelity and dramatic flair solves a very real workflow bottleneck.

The real test will be whether MAI-Voice-1 can stick to the script without sounding robotic, but on paper, giving creators verbatim control is an obvious win.