Annotate Audio Support

Teaching a new sound and it won't sit? Here are quick fixes and the most common questions.

Talk to the developer

Bug, idea or a sound it just won't learn. Every email gets read.

Troubleshooting

Train is greyed out

A model needs at least two sounds with 5 recordings each. The pup tells you how many reps are still missing. For better results, keep recording up to 50 or 100.

Training stops or fails

Keep the app open while it trains; it takes a few minutes and runs on your iPhone. If it keeps failing, email us and mention which iPhone and iOS version you use.

Auto doesn't record anything

Check that a trained model is selected under Auto recorder model, that Auto is on, and that microphone access is allowed in Settings → Privacy & Security → Microphone. Setting the confidence to LOW makes it record more easily.

It records the wrong sounds

Move the wrong clips to the “Not …” folder, add a few more reps there, and train again. Setting the confidence to HI makes it stricter.

My recordings don't show up on another device

Both devices need to be signed in to the same Apple Account with iCloud Drive turned on. Large folders can take a while to sync.

Frequently asked questions

What does Annotate Audio actually do?

It lets you train your own machine-learning sound classifier. You record labelled examples of each sound, it trains a Core ML model on your iPhone, and the model then listens live and records whenever it recognises one of your sounds.

How many examples does it need?

Five recordings per sound, for at least two sounds, is enough to train a first model. Accuracy improves with more: the app suggests 50, then 100 per sound.

Why add a “Not …” sound?

A classifier needs something to compare against. Next to “Dog”, record a “Not Dog” class with the other sounds of the room: talking, the TV, footsteps.

Does training happen in the cloud?

No. The model is trained on your iPhone with Apple’s Create ML and takes a few minutes. Afterwards you see its accuracy, and clips it disagrees with are flagged for review. Listening runs on your iPhone too.

Does it keep listening in the background?

Yes. With Auto on it keeps listening while you use other apps, and the screen dims when the phone lies face down or is covered.

Can it listen for several sounds at once?

Yes. A model can know several sounds, and each clip is filed in the folder of the sound it heard.

Can I use my data and models elsewhere?

Yes. Export any folder as a ZIP with a training layout for Create ML on Mac. Trained models are saved as Core ML files (.mlmodel) and, with iCloud Drive on, sync with your recordings.

Requirements

iPhone with iOS 18 or later. Training runs on the iPhone itself. Syncing between devices needs iCloud Drive.