v0.6.0-beta30

Released 2026-09-24 · Download .dmg (37 MB)

Name a few voices, and the app can work out the rest.

Short clips land on whoever they sound nearest

A one-second “Yeah.” barely carries a voice. There isn’t enough in it to tell who spoke, so it gets grouped with whichever voice it sits closest to — and that is often the wrong one. Clips under a second and a half are close enough to a second voice about one time in six.

Naming someone changes the situation. Once a voice has a name, the app can compare against how that person sounds across every meeting they appear in, rather than against a single clip. That is a far steadier reference.

Re-check voices

The speaker panel has a new button. It re-examines every unnamed and loosely-held segment against the people you have identified, and then splits what it finds in two:

It gets better each time you use it: name a few, re-check, name a few more. Every pass has more to go on than the last.

When two people talk at once

Some clips catch two people speaking over each other. The recording genuinely holds both voices, so no amount of re-checking can separate them — the clip isn’t close to the wrong person, it is honestly between two. Until now those were quietly filed under whoever they resembled most.

They are now marked in the segment list and sorted to the top of the re-check, where hearing two seconds settles what no measurement can.

Your corrections stay put

A segment you place by hand is treated as settled and never reconsidered, no matter how many times you re-check afterwards.

Search by who was there

Typing a name into the library search now finds meetings that person spoke in — not only meetings whose title or transcript happens to mention them.

Auto-update

Existing beta channel testers get this within 24 hours. Or trigger now: Minutiva Private → Check for Updates…