@@ -10,14 +10,154 @@ import AVFoundation
1010
1111@objc
1212public class AppleSpeechImpl : NSObject {
13+ private let speechSynthesizer = AVSpeechSynthesizer ( )
14+
1315 @objc
14- public func isAvailable( ) -> Bool {
15- return true // AVSpeechSynthesizer is available on all iOS versions we support
16+ public func getVoices( _ resolve: @escaping ( [ Any ] ) -> Void , reject: @escaping ( String , String , Error ? ) -> Void ) {
17+ Task {
18+ if #available( iOS 17 . 0 , * ) {
19+ await withCheckedContinuation { continuation in
20+ AVSpeechSynthesizer . requestPersonalVoiceAuthorization { _ in
21+ continuation. resume ( )
22+ }
23+ }
24+ }
25+
26+ let allVoices = AVSpeechSynthesisVoice . speechVoices ( )
27+ let voiceInfos = allVoices. map { $0. toDictionary ( ) }
28+ resolve ( voiceInfos)
29+ }
1630 }
1731
1832 @objc
19- public func generate( _ text: String , options: [ String : Any ] ? , resolve: @escaping ( Any ? ) -> Void , reject: @escaping ( String , String , Error ? ) -> Void ) {
20- // TODO: Implement text-to-speech functionality
21- resolve ( nil )
33+ public func generateAudio( _ text: String , options: [ String : Any ] , resolve: @escaping ( Data ) -> Void , reject: @escaping ( String , String , Error ? ) -> Void ) {
34+ let utterance = AVSpeechUtterance ( string: text)
35+
36+ utterance. voice = if let voiceId = options [ " voice " ] as? String {
37+ AVSpeechSynthesisVoice ( identifier: voiceId)
38+ } else if let language = options [ " language " ] as? String {
39+ AVSpeechSynthesisVoice ( language: language)
40+ } else {
41+ nil
42+ }
43+
44+ var collectedBuffers : [ AVAudioPCMBuffer ] = [ ]
45+
46+ var resolveCallback : ( ( Data ) -> Void ) ? = resolve
47+ var rejectCallback : ( ( String , String , Error ? ) -> Void ) ? = reject
48+
49+ speechSynthesizer. write ( utterance) { buffer in
50+ guard let pcm = buffer as? AVAudioPCMBuffer else { return }
51+
52+ if pcm. frameLength == 0 {
53+ guard let resolve = resolveCallback, let reject = rejectCallback else { return }
54+
55+ do {
56+ let data = try AppleSpeechImpl . wavData ( from: collectedBuffers)
57+ resolve ( data)
58+ } catch {
59+ reject ( " AppleSpeech " , " Error generating WAV data " , error)
60+ }
61+
62+ resolveCallback = nil
63+ rejectCallback = nil
64+ return
65+ }
66+
67+ collectedBuffers. append ( pcm)
68+ }
69+ }
70+ }
71+
72+ extension AppleSpeechImpl {
73+ /// Build a single WAV file by generating the header using the first buffer's
74+ /// format and then concatenating the raw PCM payloads of all subsequent buffers.
75+ /// Assumes all buffers share the same format and are WAV-compatible.
76+ static func wavData( from buffers: [ AVAudioPCMBuffer ] ) throws -> Data {
77+ guard let first = buffers. first else {
78+ throw NSError ( domain: " WAV " , code: - 2 ,
79+ userInfo: [ NSLocalizedDescriptionKey: " No audio buffers collected " ] )
80+ }
81+
82+ let channels = Int ( first. format. channelCount)
83+ let sampleRate = Int ( first. format. sampleRate)
84+ let isFloat32 = ( first. format. commonFormat == . pcmFormatFloat32)
85+ let bitsPerSample = isFloat32 ? 32 : 16
86+ let byteRate = sampleRate * channels * bitsPerSample / 8
87+ let blockAlign = channels * bitsPerSample / 8
88+
89+ // Helper: little-endian encoders
90+ func le16( _ v: Int ) -> [ UInt8 ] { [ UInt8 ( v & 0xff ) , UInt8 ( ( v >> 8 ) & 0xff ) ] }
91+ func le32( _ v: Int ) -> [ UInt8 ] {
92+ [ UInt8 ( v & 0xff ) , UInt8 ( ( v >> 8 ) & 0xff ) ,
93+ UInt8 ( ( v >> 16 ) & 0xff ) , UInt8 ( ( v >> 24 ) & 0xff ) ]
94+ }
95+
96+ // Estimate capacity from actual valid bytes in each buffer
97+ let estimatedCapacity = buffers. reduce ( 0 ) { acc, buf in
98+ let audioBuffer = buf. audioBufferList. pointee. mBuffers
99+ return acc + Int( audioBuffer. mDataByteSize)
100+ }
101+
102+ var payload = Data ( )
103+ payload. reserveCapacity ( estimatedCapacity)
104+
105+ // Concatenate payloads using mDataByteSize, which is kept in sync with frameLength
106+ for buf in buffers {
107+ let m = buf. audioBufferList. pointee. mBuffers
108+ let byteCount = Int ( m. mDataByteSize)
109+ if let p = m. mData {
110+ payload. append ( contentsOf: UnsafeRawBufferPointer ( start: p, count: byteCount) )
111+ }
112+ }
113+
114+ let dataChunkSize = payload. count
115+ let fmtChunkSize = 16
116+ let riffChunkSize = 4 + ( 8 + fmtChunkSize) + ( 8 + dataChunkSize)
117+
118+ var header = Data ( )
119+ header. append ( contentsOf: Array ( " RIFF " . utf8) )
120+ header. append ( contentsOf: le32 ( riffChunkSize) )
121+ header. append ( contentsOf: Array ( " WAVE " . utf8) )
122+
123+ // fmt chunk
124+ header. append ( contentsOf: Array ( " fmt " . utf8) )
125+ header. append ( contentsOf: le32 ( fmtChunkSize) )
126+ header. append ( contentsOf: le16 ( isFloat32 ? 3 : 1 ) ) // 3 = IEEE float, 1 = PCM
127+ header. append ( contentsOf: le16 ( channels) )
128+ header. append ( contentsOf: le32 ( sampleRate) )
129+ header. append ( contentsOf: le32 ( byteRate) )
130+ header. append ( contentsOf: le16 ( blockAlign) )
131+ header. append ( contentsOf: le16 ( bitsPerSample) )
132+
133+ // data chunk
134+ header. append ( contentsOf: Array ( " data " . utf8) )
135+ header. append ( contentsOf: le32 ( dataChunkSize) )
136+
137+ var out = Data ( capacity: header. count + payload. count)
138+ out. append ( header)
139+ out. append ( payload)
140+
141+ return out
142+ }
143+ }
144+
145+ extension AVSpeechSynthesisVoice {
146+ func toDictionary( ) -> [ String : Any ] {
147+ var data = [
148+ " identifier " : self . identifier,
149+ " name " : self . name,
150+ " language " : self . language,
151+ " quality " : quality,
152+ " isPersonalVoice " : false ,
153+ " isNoveltyVoice " : false
154+ ] as [ String : Any ]
155+
156+ if #available( iOS 17 . 0 , * ) {
157+ data [ " isPersonalVoice " ] = self . voiceTraits. contains ( . isPersonalVoice)
158+ data [ " isNoveltyVoice " ] = self . voiceTraits. contains ( . isNoveltyVoice)
159+ }
160+
161+ return data
22162 }
23163}
0 commit comments