Skip to content

Fix crash with high-rate audio inputs #107

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
f merged 1 commit into from
Sep 6, 2026
Merged

Fix crash with high-rate audio inputs #107

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 16 additions & 8 deletions Textream/Textream/DictationManager.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,13 +213,10 @@ class DictationManager {
return
}

let monoFormat = AVAudioFormat(
commonFormat: recordingFormat.commonFormat,
sampleRate: recordingFormat.sampleRate,
channels: 1,
interleaved: recordingFormat.isInterleaved
)
let tapFormat = recordingFormat.channelCount > 1 ? monoFormat : recordingFormat
guard let tapFormat = speechCaptureFormat(for: recordingFormat) else {
fail("Audio input format is unsupported.")
return
}

// Observe audio configuration changes
configurationChangeObserver = NotificationCenter.default.addObserver(
Expand All @@ -235,9 +232,20 @@ class DictationManager {

inputNode.removeTap(onBus: 0)

inputNode.installTap(onBus: 0, bufferSize: 1024, format: tapFormat) { [weak self] buffer, _ in
let waveformFrameInterval = AVAudioFrameCount(max(1, tapFormat.sampleRate / 20.0))
var framesSinceWaveformUpdate = waveformFrameInterval

inputNode.installTap(
onBus: 0,
bufferSize: speechCaptureBufferSize(for: tapFormat),
format: tapFormat
) { [weak self] buffer, _ in
self?.appendBuffer(buffer)

framesSinceWaveformUpdate &+= buffer.frameLength
guard framesSinceWaveformUpdate >= waveformFrameInterval else { return }
framesSinceWaveformUpdate %= waveformFrameInterval

guard let channelData = buffer.floatChannelData?[0] else { return }
let frameLength = Int(buffer.frameLength)
var sum: Float = 0
Expand Down
52 changes: 41 additions & 11 deletions Textream/Textream/SpeechRecognizer.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,27 @@ import Speech
import AVFoundation
import CoreAudio

private let waveformUpdatesPerSecond = 20.0
private let speechCaptureBuffersPerSecond = 40.0

func speechCaptureFormat(for hardwareFormat: AVAudioFormat) -> AVAudioFormat? {
guard hardwareFormat.channelCount > 1 else { return hardwareFormat }

// An input-node tap must use the hardware sample rate. AVAudioEngine can
// downmix channels here, but requesting a different rate raises an
// uncaught AVFAudio format-mismatch exception on high-rate USB devices.
return AVAudioFormat(
commonFormat: .pcmFormatFloat32,
sampleRate: hardwareFormat.sampleRate,
channels: 1,
interleaved: false
)
}

func speechCaptureBufferSize(for format: AVAudioFormat) -> AVAudioFrameCount {
AVAudioFrameCount(max(1024, format.sampleRate / speechCaptureBuffersPerSecond))
}

struct AudioInputDevice: Identifiable, Hashable {
let id: AudioDeviceID
let uid: String
Expand Down Expand Up @@ -469,16 +490,12 @@ class SpeechRecognizer {
return
}

// SFSpeechRecognizer requires mono audio. Multi-channel devices (e.g.
// RODECaster Pro II at 2ch/48kHz) cause the recognition task to silently
// return no results. Request a mono tap and let AVAudioEngine downmix.
let monoFormat = AVAudioFormat(
commonFormat: hardwareFormat.commonFormat,
sampleRate: hardwareFormat.sampleRate,
channels: 1,
interleaved: hardwareFormat.isInterleaved
)
let tapFormat = (hardwareFormat.channelCount > 1) ? monoFormat : hardwareFormat
// SFSpeechRecognizer expects mono voice audio. Downmix multi-channel
// devices without changing the input node's hardware sample rate.
guard let tapFormat = speechCaptureFormat(for: hardwareFormat) else {
failListening("Audio input format is unsupported.")
return
}

// Observe audio configuration changes (e.g. mic switched externally) to restart gracefully
configurationChangeObserver = NotificationCenter.default.addObserver(
Expand All @@ -496,9 +513,22 @@ class SpeechRecognizer {
// Belt-and-suspenders: ensure no stale tap exists before installing
inputNode.removeTap(onBus: 0)

inputNode.installTap(onBus: 0, bufferSize: 1024, format: tapFormat) { [weak self] buffer, _ in
let waveformFrameInterval = AVAudioFrameCount(
max(1, tapFormat.sampleRate / waveformUpdatesPerSecond)
)
var framesSinceWaveformUpdate = waveformFrameInterval

inputNode.installTap(
onBus: 0,
bufferSize: speechCaptureBufferSize(for: tapFormat),
format: tapFormat
) { [weak self] buffer, _ in
self?.appendBufferToRequest(buffer)

framesSinceWaveformUpdate &+= buffer.frameLength
guard framesSinceWaveformUpdate >= waveformFrameInterval else { return }
framesSinceWaveformUpdate %= waveformFrameInterval

guard let channelData = buffer.floatChannelData?[0] else { return }
let frameLength = Int(buffer.frameLength)
var sum: Float = 0
Expand Down