AI voice news, week of Oct 7, 2026: what changed for voiceover teams

Three things changed for teams that make voiceovers in the last two weeks: Microsoft added MAI-Voice-2.1 and a Flash variant, Decagon launched Voice 3 with its Chord speech model for live calls, and Google shipped Gemini 3.8 TTS, whose paid prices double on January 1, 2027. Mistral's Voxtral TTS (March 2026) is the open-weights reference point. Here is the dated list.

What happened and when

Each row is from the vendor's own page.

Actions by role

Give each person one thing to do.

• Producers: listen to one script on MAI-Voice-2.1, Gemini Flash TTS and your current voice, blind.
• Finance: add Gemini's January 1, 2027 price step and Microsoft's end-of-year intro rate to the 2027 budget.
• Developers: note that Microsoft marks the MAI voices as public preview.
• Support leads: if you run live calls, read the Voice 3 page.

Where Sume stands

Sume's TTS 1.0 is $0.0475 per 1,000 characters and speech-to-text is $0.01 per audio minute on the public rate card, with word timings and sentence slices in the result. Sume does not list the vendors above as models on its TTS surface, so use their files by importing them, or use Sume's own voices. Confirm rates in GET /v1/catalog.

What not to do

Do not switch a live client campaign to a preview voice in the week it launches. Keep the approved files, test on a new draft, and decide after a listening round.