Saj Speak — Voice Recording

Help us train a voice assistant to recognise its wake word.

You will be asked to say a short phrase into your mic several times.

Say it naturally — like you are talking to a smart speaker across the room. Do not whisper, do not shout, just your normal voice.
Vary it a little each time — sometimes faster, sometimes slower, sometimes from further away. This is what makes the training data useful.
10 recordings minimum — more is always better. 20 would be brilliant.
Phone held comfortably — about 30 cm away, not right against your mouth. Your audio is only used for model training.

It takes about 2 minutes. Your audio is only used for model training.

Your First Name Keyword

Your Environment

Say clearly:

Hey Bella

Tap to record (or press space)

0

Takes

10

Target

✓

Thank you!

Your recordings have been uploaded.

Your voice recordings were uploaded automatically and will be used to train our wake word detection model. No further action needed.