We’re building the transcription tool we wish existed.

Why we’re building this

The transcription tools that dominate the category were built for live business meetings. They’re great at that. They’re mediocre at almost everything else: 8-speaker panels, longitudinal research, accent-heavy interviews, code-switched conversations, recordings where the same person needs to stay named across years of files.

The unsexy detail every other tool gets wrong is speaker identity. You upload a recording, you get back “Speaker 1, Speaker 2,” and the same person is “Speaker 3” in the next file. For the people who do this work full-time, the cleanup tax adds up to days per year. The transcription part is solved. The labels on the words are where the value leaks.

We’re building a transcription product where speaker identity is the core feature, not a checkbox. Persistent voice IDs across files, smart auto-labeling from transcript context, honest privacy defaults, no bill shock.

What we’re not going to do

A short, deliberate list. We won’t train models on your audio. We won’t join your meetings as an unannounced bot. We won’t add a per-minute meter that double-counts edit replays. We won’t silently change retention defaults to accumulate more data. We won’t pitch the legal department on enterprise SaaS while quietly rewriting our terms to expand what we can do with your recordings.

If we ever do any of those things, this page will say so first.