Here is the target: SoundSense is designed to stay well under 10 percent of your iPhone's battery per hour of active listening. That is the engineering bar we set before we started writing code, and it is the one we want you to know before you install the app.
Context for the number
Under 10 percent per hour sounds like a lot or a little depending on what you are comparing it to. For reference, approximate hourly drain on a modern iPhone in normal conditions:
- Streaming music (Spotify, screen off, Bluetooth): about 4 to 8 percent per hour
- GPS turn-by-turn navigation (screen on): about 15 to 20 percent per hour
- Active video call (FaceTime, screen on): about 20 to 25 percent per hour
- Idle in your pocket (no apps running): under 1 percent per hour
So SoundSense in active mode aims to land roughly in the range of a music streaming session. More than idle, less than navigation. That is where an always-on audio classifier with a well-engineered pipeline should land, and it is where we are building toward.
How we got there
The naive way to build a sound awareness app is to run the full classifier on every incoming audio frame, 24/7. That is simple code and it would burn 20 to 30 percent of your battery per hour. Unacceptable.
Instead, SoundSense uses a layered detection pipeline. Stage one is a very lightweight energy-and-frequency monitor that looks at incoming audio and decides whether anything interesting is happening. If not, and most of the time the answer is "no, the room is quiet," the full classifier never wakes up. This stage uses a rounding-error amount of power.
Stage two is a fast frequency-band match that runs when stage one finds a sharp onset. This is fast enough that it adds minimal drain even when triggering often.
Stage three, the full classifier, only runs when stages one and two agree that something worth classifying is in the audio. In a typical home, this might run only a few dozen times an hour. In a loud environment like a busy cafe or a construction site, it will run more often, and the app's battery cost will push toward the top of its target range.
The practical upshot: SoundSense is designed to be efficient in quiet environments (where it mostly has nothing to do) and reasonable in loud environments (where it has to work harder). Either way, the goal is that you are not going to lose a day of battery to it.
Low-power mode
When your iPhone is in low-power mode, SoundSense respects it. In low-power mode we drop the frame rate of the energy monitor, widen the thresholds, and reduce classifier invocations. Detection quality degrades. You will miss some edge cases, but the power cost drops significantly. We think of this as a safety-net mode rather than normal operation, but it is available when you need it.
Practical patterns
Charge while you sleep. The most reliable way to keep a full day of SoundSense is to charge overnight. Starting at 100 percent in the morning with SoundSense running all day should leave comfortable headroom for the evening.
Top up around lunch. If you are out all day, a short top-up around midday will comfortably get you to bedtime.
Keep wireless earbuds charged. Bluetooth audio output to earbuds or hearing aids is a small additional draw. Not huge, but worth knowing about.
Apple Watch is coming later. A dedicated Watch app is on the roadmap as part of SoundSense+. When it ships, you will be able to route primary alerts to your wrist and keep the phone screen off more of the time.
The commitment
We are going to keep publishing honest numbers here, including when they get worse. If a future version of SoundSense pushes above its target range because we added a feature that is not worth the cost, you should be the first to know, and we should probably roll it back.
Under 10 percent per hour of active listening is the target. Anything above that, we will earn it by being honest about why.