|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Cion\TextToSpeech\Converters; |
|
4
|
|
|
|
|
5
|
|
|
use Aws\Polly\PollyClient; |
|
6
|
|
|
use Aws\Result; |
|
7
|
|
|
use Cion\TextToSpeech\Contracts\Converter; |
|
8
|
|
|
use Cion\TextToSpeech\Traits\HasLanguage; |
|
9
|
|
|
use Cion\TextToSpeech\Traits\Sourceable; |
|
10
|
|
|
use Cion\TextToSpeech\Traits\Storable; |
|
11
|
|
|
use Illuminate\Support\Arr; |
|
12
|
|
|
|
|
13
|
|
|
class PollyConverter implements Converter |
|
14
|
|
|
{ |
|
15
|
|
|
use Storable, Sourceable, HasLanguage; |
|
16
|
|
|
|
|
17
|
|
|
/** |
|
18
|
|
|
* Client instance of Polly. |
|
19
|
|
|
* |
|
20
|
|
|
* @var \Aws\Polly\PollyClient |
|
21
|
|
|
*/ |
|
22
|
|
|
protected $client; |
|
23
|
|
|
|
|
24
|
|
|
/** |
|
25
|
|
|
* Construct converter. |
|
26
|
|
|
* |
|
27
|
|
|
* @param PollyClient $client |
|
28
|
|
|
*/ |
|
29
|
|
|
public function __construct(PollyClient $client) |
|
30
|
|
|
{ |
|
31
|
|
|
$this->client = $client; |
|
32
|
|
|
} |
|
33
|
|
|
|
|
34
|
|
|
/** |
|
35
|
|
|
* Get the Polly Client. |
|
36
|
|
|
* |
|
37
|
|
|
* @return \Aws\Polly\PollyClient |
|
38
|
|
|
*/ |
|
39
|
|
|
public function getClient(): PollyClient |
|
40
|
|
|
{ |
|
41
|
|
|
return $this->client; |
|
42
|
|
|
} |
|
43
|
|
|
|
|
44
|
|
|
/** |
|
45
|
|
|
* Converts the text to speech. |
|
46
|
|
|
* |
|
47
|
|
|
* @param string $data |
|
48
|
|
|
* @param array $options |
|
49
|
|
|
* @return string |
|
50
|
|
|
*/ |
|
51
|
|
|
public function convert(string $data, array $options = null) |
|
52
|
|
|
{ |
|
53
|
|
|
$text = $this->getTextFromSource($data); |
|
54
|
|
|
|
|
55
|
|
|
if ($this->isTextAboveLimit($text)) { |
|
56
|
|
|
$text = $this->getChunkText($text); |
|
57
|
|
|
} |
|
58
|
|
|
|
|
59
|
|
|
$result = $this->synthesizeSpeech($text, $options); |
|
60
|
|
|
|
|
61
|
|
|
if ($result instanceof Result) { |
|
62
|
|
|
// Store audio file to disk |
|
63
|
|
|
return $this->store( |
|
64
|
|
|
$this->getTextFromSource($data), |
|
65
|
|
|
$this->getResultContent($result) |
|
66
|
|
|
); |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
return $this->store( |
|
70
|
|
|
$this->getTextFromSource($data), |
|
71
|
|
|
$this->mergeOutputs($result) |
|
72
|
|
|
); |
|
73
|
|
|
} |
|
74
|
|
|
|
|
75
|
|
|
/** |
|
76
|
|
|
* Request to Amazon Polly to synthesize speech. |
|
77
|
|
|
* |
|
78
|
|
|
* @param string|array $text |
|
79
|
|
|
* @param array $options |
|
80
|
|
|
* @return array|\Aws\Result |
|
81
|
|
|
*/ |
|
82
|
|
|
protected function synthesizeSpeech($text, array $options = null) |
|
83
|
|
|
{ |
|
84
|
|
|
if (is_string($text)) { |
|
85
|
|
|
return $this->client->synthesizeSpeech([ |
|
86
|
|
|
'LanguageCode' => $this->getLanguage(), |
|
87
|
|
|
'VoiceId' => $this->voice($options), |
|
88
|
|
|
'OutputFormat' => $this->format($options), |
|
89
|
|
|
'TextType' => 'ssml', |
|
90
|
|
|
'Text' => '<speak>'.$text.'</speak>', |
|
91
|
|
|
]); |
|
92
|
|
|
} |
|
93
|
|
|
|
|
94
|
|
|
$results = []; |
|
95
|
|
|
|
|
96
|
|
|
foreach ($text as $item) { |
|
97
|
|
|
$result = $this->client->synthesizeSpeech([ |
|
98
|
|
|
'LanguageCode' => $this->getLanguage(), |
|
99
|
|
|
'VoiceId' => $this->voice($options), |
|
100
|
|
|
'OutputFormat' => $this->format($options), |
|
101
|
|
|
'Text' => $item, |
|
102
|
|
|
]); |
|
103
|
|
|
|
|
104
|
|
|
array_push($results, $result); |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
|
|
return $results; |
|
108
|
|
|
} |
|
109
|
|
|
|
|
110
|
|
|
/** |
|
111
|
|
|
* Merges the output from amazon polly. |
|
112
|
|
|
* |
|
113
|
|
|
* @return mixed |
|
114
|
|
|
*/ |
|
115
|
|
|
protected function mergeOutputs(array $results) |
|
116
|
|
|
{ |
|
117
|
|
|
$mergedResult = null; |
|
118
|
|
|
foreach ($results as $result) { |
|
119
|
|
|
$mergedResult .= $this->getResultContent($result); |
|
120
|
|
|
} |
|
121
|
|
|
|
|
122
|
|
|
return $mergedResult; |
|
123
|
|
|
} |
|
124
|
|
|
|
|
125
|
|
|
/** |
|
126
|
|
|
* Checks the length of the text if more than 3000. |
|
127
|
|
|
* |
|
128
|
|
|
* @param string $text |
|
129
|
|
|
* @return bool |
|
130
|
|
|
*/ |
|
131
|
|
|
protected function isTextAboveLimit(string $text) |
|
132
|
|
|
{ |
|
133
|
|
|
return strlen($text) > 2000; |
|
134
|
|
|
} |
|
135
|
|
|
|
|
136
|
|
|
/** |
|
137
|
|
|
* Chunk the given text into array. |
|
138
|
|
|
* |
|
139
|
|
|
* @param string $text |
|
140
|
|
|
* @param int $size |
|
141
|
|
|
* @return array |
|
142
|
|
|
*/ |
|
143
|
|
|
protected function getChunkText(string $text, int $size = 2000) |
|
144
|
|
|
{ |
|
145
|
|
|
return explode("\n", wordwrap($text, $size)); |
|
146
|
|
|
} |
|
147
|
|
|
|
|
148
|
|
|
/** |
|
149
|
|
|
* Get the text to speech voice ID. |
|
150
|
|
|
* |
|
151
|
|
|
* @param array $options |
|
152
|
|
|
* @return string |
|
153
|
|
|
*/ |
|
154
|
|
|
protected function voice($options) |
|
155
|
|
|
{ |
|
156
|
|
|
$default = config('tts.services.polly.voice_id', 'Amy'); |
|
157
|
|
|
|
|
158
|
|
|
return Arr::get($options, 'voice', $default); |
|
159
|
|
|
} |
|
160
|
|
|
|
|
161
|
|
|
/** |
|
162
|
|
|
* Get the language |
|
163
|
|
|
* |
|
164
|
|
|
* @return string |
|
165
|
|
|
*/ |
|
166
|
|
|
protected function getLanguage() |
|
167
|
|
|
{ |
|
168
|
|
|
return $this->language ?? config('tts.language', 'en-US'); |
|
169
|
|
|
} |
|
170
|
|
|
|
|
171
|
|
|
/** |
|
172
|
|
|
* Get the audio format. |
|
173
|
|
|
* |
|
174
|
|
|
* @param array $options |
|
175
|
|
|
* @return string |
|
176
|
|
|
*/ |
|
177
|
|
|
protected function format($options) |
|
178
|
|
|
{ |
|
179
|
|
|
$default = config('tts.output_format', 'mp3'); |
|
180
|
|
|
|
|
181
|
|
|
return Arr::get($options, 'format', $default); |
|
182
|
|
|
} |
|
183
|
|
|
|
|
184
|
|
|
/** |
|
185
|
|
|
* Get the content of the result from AWS Polly. |
|
186
|
|
|
* |
|
187
|
|
|
* @param mixed $result |
|
188
|
|
|
* @return mixed |
|
189
|
|
|
*/ |
|
190
|
|
|
protected function getResultContent($result) |
|
191
|
|
|
{ |
|
192
|
|
|
return $result->get('AudioStream')->getContents(); |
|
193
|
|
|
} |
|
194
|
|
|
} |
|
195
|
|
|
|