Recording and speech to text
Recording ↑
Press Record. The first time, the app asks where recordings go (on a Mac, Music/KaikuSDR unless you choose another folder)
and how much room they may take; then what happens when that is full: delete the oldest, or stop recording. Every transmission
becomes a sound file, named in the activity log with its channel and time.
Recording radio communications may be restricted or prohibited where you are. You are responsible for deciding whether to record, and for complying with the laws that apply to you.
A held channel can also record the receiver's whole signal (IQ), to play back later as if it were a receiver.
Playing recordings ↑
The Recordings tab under the chart lists what was kept, by day: the time each transmission began, the channel, the band, how long it was, which receiver heard it, and the words when speech to text wrote them down. A row where somebody talked carries the same speech mark as Heard recently; a row without one was a bare carrier, a tone, noise or data, as far as the speech detector could tell. Only where somebody talked hides the rest. Press the play button on a row to hear it again through the app's own sound output while scanning carries on; the row says how far it has got, and the same button stops it. The button beside play, play from here on, plays that transmission and everything heard after it, in order, so that you can follow how a conversation went; the row says how many are still to come, and Next moves on. The box beside the day narrows the list by name, frequency or words. Open the folder shows the files themselves.
The bin on a row deletes that recording, after asking: the file is removed from the folder for good, and the activity log keeps the transmission but no longer names a file for it. Delete the … where nobody talked does the same for every recording of the day that speech detection marked as no talk, in one go, after asking; recordings the app has no record of (copied in, or made by an older version) are kept, since nothing says what they hold.
Each file carries what it is: any player that shows a file's information shows the title, when it began and the channel's name, and the app reads the rest from the file too, so a recording copied elsewhere and back is still known.
Speech to text ↑
Transcribe what is said (in Settings, on the Speech to text page) turns speech into text, on your computer: nothing is sent anywhere. It needs a model, downloaded once when you choose it:
- Balanced: good. For English.
- Best: slow; the one for other languages, Finnish among them.
Downloading a model turns speech to text on once the model is here and checked. What was said appears under Conversation, and in the activity log.
On a slow computer the words may come later than the talk, and when too many transmissions are waiting the oldest are left out: the Speech to text page says how many, and the row in Conversation reads "not written down". A smaller model is quicker.