How it works
Where your recording goes, and how the results are checked.
From your recording to the page
Your answer is recorded in the browser and sent to Sayfield's server.
Open models write down the words and place each one in time.
Three separate detectors look for pauses; a pause counts when two of them agree.
Sayfield's own rules mark the fillers and group the words into phrases. Pitch is measured with Praat, a phonetics tool, but not shown in the app yet.
What it's built on
First came a review of the research: how unscripted speech is transcribed word for word, how pauses and phrase endings are marked, and what feedback on speaking helps.
One finding shaped the product: speech recognisers drop fillers, and in our check they dropped more for second-language speakers. Keeping them is the first job for Sayfield's models.
How it's checked
Checked today. Today's words and fillers were checked on public meeting recordings transcribed by hand. The app still misses some fillers, more often for second-language speakers.
Tested next. Sayfield's own models will be tested on recordings they never saw, against marks checked by people, never another machine's raw output. Results will be split by first language, and no figure is published before a test backs it.
The notes
In developmentShort notes will quote your own words back to you: one thing to keep, one move to try, and where the move comes from.
An AI model will write them from your transcript, following Sayfield's speaking criteria. They can be wrong, so every claim quotes your words for you to check.
No accent score. Ever.
No grade for you as a speaker, no labels for confidence or emotion. Sayfield describes what was said and how, never the person who said it.
What's kept and who sees it: Privacy