How it works
Tap ▶ to play. Use the video controls to pause, seek or expand.
Listening to M0MCX with QSO Transcriber
Listening to a real amateur radio recording from M0MCX inside QSO Transcriber.
In this demonstration, you can see how the app listens to the radio audio, converts the speech into text and automatically detects and collects callsigns from the conversation.
Real-world radio audio is very different from clean studio speech. Noise, fading, interference, accents and varying signal strength can occasionally challenge Apple’s Speech-to-Text engine, so not every spoken word will always be transcribed perfectly.
This is where the QSO Context Engine comes into play.
QSO Transcriber analyses the transcription in the context of amateur radio communication and looks for information such as callsigns and other useful QSO data. In the video you can see detected callsigns being identified and collected automatically while the recording is playing.
The goal is to make it easier to follow a QSO, catch important information and review the conversation afterwards — a little like having subtitles for amateur radio.
Tap ▶ to play. Use the video controls to pause, seek or expand.
Importing a QSO recording
I’m using an existing QSO recording here to demonstrate import, transcription and checking the log details.
An imported recording goes through the same speech recognition and local QSO analysis as audio recorded in the app. The original recording remains available for comparison.
After importing the recording, you can:
- Listen to the original QSO recording
- Read the processed transcript alongside the audio
- Check suggested callsigns, signal reports and QTH details
- Use timestamps to replay an uncertain passage
- Correct the proposed log fields without changing the original audio or raw transcript
- Save the reviewed contact with its recording and transcript in the Logbook
Tap ▶ to play. Use the video controls to pause, seek or expand.
Live transcription
In this recording, QSO Transcriber listens to audio from a KiwiSDR receiver and transcribes the received speech as it arrives. Speech recognition can mishear callsigns and phonetics, so the text still needs checking against the audio.
Audio processing and log review
Recorded or imported audio can pass through optional noise reduction before speech recognition. The QSO Context Engine analyses the resulting text locally and proposes log fields. You check the callsign, report and other details before saving or exporting the contact. The original audio and raw transcript are kept separately from your corrections.
Developing the QSO Context Engine
I use amateur-radio recordings to compare the received audio with the transcript and proposed log fields. QRM, QSB, accents and variations in phonetics all make this work less straightforward than ordinary speech transcription. I check the results by hand and use the errors to guide changes, then test again. There is still work to do; the engine’s output is a set of suggestions for the operator to review.
