jake/vOOice
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix audio processing overflow error

Browse Source
This commit is contained in:
Beingpax 2025-08-14 14:22:28 +05:45
parent f909185eac
commit 36fa030d12
1 changed files with 54 additions and 87 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

141
VoiceInk/Services/AudioFileProcessor.swift
View File

@ -34,26 +34,16 @@ class AudioProcessor {
}
}
/// Process audio file and return samples ready for Whisper
/// - Parameter url: URL of the input audio file
/// - Returns: Array of normalized float samples
func processAudioToSamples(_ url: URL) async throws -> [Float] {
logger.notice("🎵 Processing audio file to samples: \(url.lastPathComponent)")
// Create AVAudioFile from input
guard let audioFile = try? AVAudioFile(forReading: url) else {
logger.error("❌ Failed to create AVAudioFile from input")
throw AudioProcessingError.invalidAudioFile
}
// Get format information
let format = audioFile.processingFormat
let sampleRate = format.sampleRate
let channels = format.channelCount
let totalFrames = audioFile.length
logger.notice("📊 Input format - Sample Rate: \(sampleRate), Channels: \(channels)")
// Create output format (always 16kHz mono float)
let outputFormat = AVAudioFormat(
commonFormat: .pcmFormatFloat32,
sampleRate: AudioFormat.targetSampleRate,
@ -62,76 +52,69 @@ class AudioProcessor {
)
guard let outputFormat = outputFormat else {
logger.error("❌ Failed to create output format")
throw AudioProcessingError.unsupportedFormat
}
// Read input file into buffer
let inputBuffer = AVAudioPCMBuffer(
pcmFormat: format,
frameCapacity: AVAudioFrameCount(audioFile.length)
)
let chunkSize: AVAudioFrameCount = 50_000_000
var allSamples: [Float] = []
var currentFrame: AVAudioFramePosition = 0
guard let inputBuffer = inputBuffer else {
logger.error("❌ Failed to create input buffer")
throw AudioProcessingError.conversionFailed
}
try audioFile.read(into: inputBuffer)
// If format matches our target, just convert to samples
if sampleRate == AudioFormat.targetSampleRate && channels == AudioFormat.targetChannels {
logger.notice("✅ Audio format already matches requirements")
return convertToWhisperFormat(inputBuffer)
}
// Create converter for format conversion
guard let converter = AVAudioConverter(from: format, to: outputFormat) else {
logger.error("❌ Failed to create audio converter")
throw AudioProcessingError.conversionFailed
}
// Create output buffer
let ratio = AudioFormat.targetSampleRate / sampleRate
let outputBuffer = AVAudioPCMBuffer(
pcmFormat: outputFormat,
frameCapacity: AVAudioFrameCount(Double(inputBuffer.frameLength) * ratio)
)
guard let outputBuffer = outputBuffer else {
logger.error("❌ Failed to create output buffer")
throw AudioProcessingError.conversionFailed
}
// Perform conversion
var error: NSError?
let status = converter.convert(
to: outputBuffer,
error: &error,
withInputFrom: { inNumPackets, outStatus in
outStatus.pointee = .haveData
return inputBuffer
while currentFrame < totalFrames {
let remainingFrames = totalFrames - currentFrame
let framesToRead = min(chunkSize, AVAudioFrameCount(remainingFrames))
guard let inputBuffer = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: framesToRead) else {
throw AudioProcessingError.conversionFailed
}
)
if let error = error {
logger.error("❌ Conversion failed: \(error.localizedDescription)")
throw AudioProcessingError.conversionFailed
audioFile.framePosition = currentFrame
try audioFile.read(into: inputBuffer, frameCount: framesToRead)
if sampleRate == AudioFormat.targetSampleRate && channels == AudioFormat.targetChannels {
let chunkSamples = convertToWhisperFormat(inputBuffer)
allSamples.append(contentsOf: chunkSamples)
} else {
guard let converter = AVAudioConverter(from: format, to: outputFormat) else {
throw AudioProcessingError.conversionFailed
}
let ratio = AudioFormat.targetSampleRate / sampleRate
let outputFrameCount = AVAudioFrameCount(Double(inputBuffer.frameLength) * ratio)
guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: outputFormat, frameCapacity: outputFrameCount) else {
throw AudioProcessingError.conversionFailed
}
var error: NSError?
let status = converter.convert(
to: outputBuffer,
error: &error,
withInputFrom: { inNumPackets, outStatus in
outStatus.pointee = .haveData
return inputBuffer
}
)
if let error = error {
throw AudioProcessingError.conversionFailed
}
if status == .error {
throw AudioProcessingError.conversionFailed
}
let chunkSamples = convertToWhisperFormat(outputBuffer)
allSamples.append(contentsOf: chunkSamples)
}
currentFrame += AVAudioFramePosition(framesToRead)
}
if status == .error {
logger.error("❌ Conversion failed with status: error")
throw AudioProcessingError.conversionFailed
}
logger.notice("✅ Successfully converted audio format")
return convertToWhisperFormat(outputBuffer)
return allSamples
}
/// Convert audio buffer to Whisper-compatible samples
private func convertToWhisperFormat(_ buffer: AVAudioPCMBuffer) -> [Float] {
guard let channelData = buffer.floatChannelData else {
logger.error("❌ No channel data available in buffer")
return []
}
@ -139,16 +122,9 @@ class AudioProcessor {
let frameLength = Int(buffer.frameLength)
var samples = Array(repeating: Float(0), count: frameLength)
logger.notice("📊 Converting buffer - Channels: \(channelCount), Frames: \(frameLength)")
// If mono, just copy the samples
if channelCount == 1 {
samples = Array(UnsafeBufferPointer(start: channelData[0], count: frameLength))
logger.notice("✅ Copied mono samples directly")
}
// If stereo or more, average all channels
else {
logger.notice("🔄 Converting \(channelCount) channels to mono")
} else {
for frame in 0..<frameLength {
var sum: Float = 0
for channel in 0..<channelCount {
@ -158,19 +134,11 @@ class AudioProcessor {
}
}
// Normalize samples to [-1, 1]
let maxSample = samples.map(abs).max() ?? 1
if maxSample > 0 {
logger.notice("📈 Normalizing samples with max amplitude: \(maxSample)")
samples = samples.map { $0 / maxSample }
}
// Log sample statistics
if let min = samples.min(), let max = samples.max() {
logger.notice("📊 Final sample range: [\(min), \(max)]")
}
logger.notice("✅ Successfully converted \(samples.count) samples")
return samples
}
func saveSamplesAsWav(samples: [Float], to url: URL) throws {
@ -213,7 +181,6 @@ class AudioProcessor {
)
try audioFile.write(from: buffer)
logger.notice("✅ Successfully saved processed audio to \(url.lastPathComponent)")
}
}