No bars, no problem. Recount converts speech to text on your own laptop, so it does not care whether the property has signal, and neither do you.
Most dictation tools send your audio to a server. That is fine at a desk and useless in a basement, a crawlspace, an attic, or a rural property twenty minutes off the highway. You speak, nothing happens, and you end up typing the findings twice — once into your phone's notes, again into the report that evening.
Recount has no server to reach. The speech model is inside the program, running on your processor. Turn off your wifi and it behaves identically.
General dictation software transcribes inspection terms badly, because nothing in its training weighted them. You say condensate and get "condense eight". You say efflorescence and get something unprintable.
Recount ships an inspection vocabulary you switch on from the toolbar. Measured on a test set of inspection phrasing, it cut word errors by a fifth.
Walking a client through the property while recording? Turn on speaker labelling and Recount separates the voices — and recognises you again when you start talking after the client has finished, rather than just noting that somebody changed.
Speaker 1: The north elevation shows efflorescence along the foundation. Speaker 2: Is that something we need to worry about? Speaker 1: It indicates moisture moving through the masonry.
It clusters voices, it does not know names — whoever speaks first is Speaker 1. People talking over each other get assigned to one of them, not both.
| Typical dictation subscription | Recount | |
|---|---|---|
| Price | $7–15/month | $29.99 once |
| Three years | $252–540 | $29.99 |
| Works with no signal | No | Yes |
| Inspection vocabulary | Add your own | Built in |
It is a young product and there are things it does not do:
No card. Run it on a real inspection and see whether it handles your vocabulary and your report software. If it does not, you have lost an afternoon rather than a subscription.