FDSoundActivatedRecorder

FDSoundActivatedRecorder

 * V
 * O                 /-----------\
 * L                /             \
 * U          Rise /               \ Fall
 * M              /                 \
 * E  -----------/                   \-----------
 *
 *       Quiet   |   Recorded file   |   Quiet

Example

Clone the repo and open Example/Example.xcodeproj. Run the Example scheme on a recent iPhone simulator.

Installation

Add this package with Swift Package Manager. In Xcode that is File > Add Package Dependencies…

Development

Run the package tests on an iPhone simulator. Package.swift requires iOS 16 and the library records through AVFoundation, so the destination is a simulator. This is the same command .github/workflows/ci.yml runs, with the simulator chosen from the ones installed on this Mac:

UDID=$(xcrun simctl list devices available | awk -F '[()]' '/iPhone/ { print $2; exit }')
xcodebuild -scheme FDSoundActivatedRecorder -destination "id=$UDID" test

Maintenance and dependency updates

Do this every quarter, and send a pull request when an update is safe. The list is the one in swift-module-template, applied to the files this repository actually has:

1. Review the external Actions in .github/workflows/ci.yml and the xcode-27 runner image. GitHub-supported Actions (the actions/ organization) need a short review.
2. Re-read swift-module-template at the release in References. When a newer release is published, adopt it and update that citation in the same change.

License

FDSoundActivatedRecorder is available under the MIT license. See the LICENSE file for more information.

Features

• You can start recording when sound is detected
• Sound stops recording when the user is done talking
• Works with ARC and iOS 16+

Usage

Import the project with:

import FDSoundActivatedRecorder

Then begin listening with:

self.recorder = FDSoundActivatedRecorder()
self.recorder.delegate = self
self.recorder.startListening()

A full implementation example is provided in this project.

Regular recorder

If you want to use it as a regular recorder, without trimming the audio, you can tweak the Configuration to achieve that.

Full API

The full API, from FDSoundActivatedRecorder.swift is copied below:

// FIXME: PASTE HERE

Technical discussion

This library is tuned for human speech detection using Apple retail iOS devices in a quiet or noisy environement. You are welcome to tune the audio detection constants of this program for any special needs you may have. Following is a technical description of how the algorithm works from FDSoundActivatedRecorder.swift.

 * V
 * O                 /-----------\
 * L                /             \
 * U          Rise /               \ Fall
 * M              /                 \
 * E  -----------/                   \-----------
 *
 *       Quiet   |   Recorded file   |   Quiet

• We listen and save audio levels every INTERVAL
• When several levels exceed the recent moving average by a threshold, we record
• (The exceeding levels are not included in the moving average)
• When several levels deceed the recent moving average by a threshold, we stop recording
• (The deceeding levels are not included in the moving average)

References

1. The Example app is one application target, package tests use Swift Testing, and the Example, Installation, Development, and maintenance sections follow swift-module-template release v16.6.0 (2026-10-12), commit 5c2368b.
2. From that release, .github/workflows/ci.yml runs on the GitHub-hosted xcode-27 image and checks out with actions/checkout@v7. The template’s macOS job is .github/workflows/swiftlang-workflows.yml. macos-26 stops at Xcode 26. The test command stays xcodebuild against an iPhone simulator, because this package imports AVFoundation and declares .iOS(.v16).
3. Deltas from that release, kept because this library records on iOS:
   • Package.swift stays swift-tools-version: 6.0 and .iOS(.v16). The library does not enable ApproachableConcurrency. The Example app stays on iOS 16.6. The template package is swift-tools-version: 6.4 with that upcoming feature, and it has no platform restriction.
   • .github/workflows/build-test.yml builds a release-mode Linux library and .github/workflows/release.yml publishes it with Release Please. This repository keeps the changelog and the release steps in CONTRIBUTING.md. This repository does not publish the template’s Linux static library. Swift Package Manager installs this library from the git tag.
   • The template also builds Linux, Windows, Wasm, and Embedded Swift. Those jobs do not build this package. Soundness in that workflow expects a Swift 6.4 image. This repository does not run it.
   • .gitignore already inlines Swift.gitignore, which is the ignore file that release vendors.
   • lint.yml, .editorconfig, .markdownlint.json, .markdownlintignore, and .prettierignore follow swift-module-template release v16.6.0. Prettier checks files that are not Markdown. markdownlint checks Markdown. The commands are in that template’s Development section.