Models · 1 October 2026 · 1 min read
Decagon Chord: a voice model for customer calls
Announced on 1 October 2026: a voice model that slows down for a confirmation code and switches language when the caller does.
By The Crewnema team
Decagon announced Chord on 1 October 2026, its first voice model, as part of its Voice 3 system for customer-service calls. These are the points from Decagon’s announcement.
What Decagon announced
- Trained for customer conversations, adjusting phrase by phrase, for example slowing down to read out a confirmation code.
- More than 70 languages, detecting the caller’s language and switching automatically, even in the middle of a sentence.
- Locale-specific voices, with brand control of voice choice and pronunciation.
- Trained on licensed data and consenting voice talent, not on customers’ data, by Decagon’s account.
- Sold as part of Decagon’s platform; no separate pricing was published.
Our view
Slowing down for a code is a small detail that shows the right idea: speech should change with what is being said, not run at one speed. The same is true of an advert, where the price and the website need to be heard clearly. Switching language mid-sentence is how many Indian customers actually speak. But this is an enterprise product for support teams; a small business will meet it, if at all, as a company it calls.
Side by side
In Crewnema
Chord is not on Crewnema’s roster.
Sources
Make your first video
Crewnema is in private beta and lets businesses in a few at a time. Request access and we will write when yours is open.