Annotate Audio

Teach your iPhone to recognise any sound. Then let it listen for you, recording only when that sound happens.

Annotate Audio learns the sounds you care about: your dog barking, a snore, a cough, a baby crying, a doorbell. Leave it listening overnight or while you're out, and come back to clips of exactly what you asked for, sorted into folders. No hours of silence to scroll through.

HOW IT WORKS
1. Pick a sound. Each sound gets its own folder. Add a "Not …" folder too, so it learns the difference.
2. Record some reps. Every recording is a rep. Five reps of two sounds is enough to start.
3. Tap Train. Your model is built right on your iPhone in a few minutes. No cloud, no code.
4. Turn on auto record. It listens and saves a clip every time it hears your sound.

NEVER MISS THE START
Auto record keeps a short rolling buffer, so every clip begins before the sound did. Manual recordings can include the 2 seconds before you tapped record, too.

A DOG THAT LEARNS WITH YOU
Your pixel pup levels up as you record: Puppy, Learning, Good dog, Top dog. It earns a collar at 5 reps, a tag at 50 and a crown at 100, and tells you what it still needs before it can train.

MADE FOR REAL LIFE
• Barking or whining while you're away
• Snoring, sleep talking and coughing at night
• A baby crying in the next room
• Doorbells, alarms, appliances and machines
• Birds, wildlife and field recordings
• Any sound you want to catch, count or collect

FEATURES
• Train custom sound classifiers on your iPhone
• Auto record triggered by the sounds you trained
• Detects several sounds at once and files each clip in its own folder
• Low, mid and high confidence settings
• Pre-buffer so clips start before the sound
• Manual recording, open-ended or 5 seconds
• Live waveform and clip playback
• Move clips between folders to clean up your training data
• Switch between several trained models
• Keeps listening in the background, and dims the screen when the phone is face down or covered
• iCloud sync of recordings and models across your devices
• ZIP export of recordings, including a training layout for Create ML on Mac

PRIVATE BY DESIGN
Training and listening happen on your device. No account. No ads. No tracking. Your recordings stay in your own iCloud.

ONE PRICE
Pay once. No subscription.

FOR MAKERS AND RESEARCHERS
Annotate Audio is also a quick way to collect and label audio datasets in the field and train Core ML sound models without writing code: bioacoustics, machine monitoring, smart-home experiments, or checking whether a sound is recognisable before you build.