What actually counts as an “hour”
You should be able to predict your bill without a spreadsheet. That means being precise about the one unit that actually shows up on it: the source-hour.
Usage follows detected speech, with a floor
We meter each live source by the hour, and usage follows the speech we actually detect and translate. There's one wrinkle: a live session holds translation capacity open the whole time it runs, so it bills for at least half of its active time. In practice, a one-hour live session uses at least 30 minutes and never more than 60.
- Counts: detected speech while translation is live, and at minimum half of the session's active time.
- Doesn't count: setup and prewarm before the translator is ready. The clock starts when translation goes live.
- Never counts: how many people are listening, or how many languages you run within your plan.
Why we measure it this way
The work of turning speech into text, text into other languages, and language back into natural audio runs from the moment a session goes live, even through quiet stretches, because that capacity stays reserved for you. The 50% floor keeps a live-but-quiet session fair without ever charging you a full hour of silence. And a room somebody forgot to close doesn't run up a bill: after 10 minutes of silence a visible countdown starts, at 15 the session ends itself, and every session has a hard 12-hour ceiling.
A good turnout should never raise your bill, and a forgotten session should never run one up.
Listeners stay free, every language you add within your plan is included, and the meter is tied to live translation time with a 50% floor. That's the whole meter.
Free listeners on every plan. Try it yourself.
Early access: be among the first when we open.
Listeners are never metered per person; very large audiences follow fair-use capacity limits. See Terms.