Merge pull request #26 from meemalabs/feature/add-support-for-speech-…

…marks

Added support for speech marks.

README.md

@@ -53,6 +53,10 @@ $path = TextToSpeech::language('ja-JP')
 // use the SSML text type in the convert() method.
 $path = TextToSpeech::ssml()
     ->convert('<speak>Hi There <break /> This is SSML syntax</speak');
+    
+// using speech marks.
+$output = TextToSpeech::speechMarks(['sentence', 'word', 'viseme', 'ssml'])
+    ->convert('This is a test'); // This will return an array.
 ```
 
 ## Installation
@@ -159,7 +163,7 @@ return [
         'polly' => [
             /**
              * Voice ID to use for the synthesis.
-             * You may use any of the following:
+             * You may use any of the following:.
              *
              * Aditi, Amy, Astrid, Bianca, Brian, Camila, Carla, Carmen, Celine,
              * Chantal, Conchita, Cristiano, Dora, Emma, Enrique, Ewa, Filiz,
@@ -171,6 +175,14 @@ return [
              */
             'voice_id' => env('AWS_VOICE_ID', 'Amy'),
 
+            /**
+             * You can request any or all of the speech mark types, but leave it empty if you don't use speech marks.
+             * You may add any of the following:.
+             *
+             * sentence, word, viseme, ssml
+             */
+            'speech_marks' => [],
+
             /**
              * IAM Credentials from AWS.
              */

composer.json

@@ -28,7 +28,8 @@
         "php": "^7.3|^7.4|^8.0",
         "aws/aws-sdk-php": "^3.178",
         "guzzlehttp/guzzle": "^6.0|^7.0",
-        "illuminate/support": "5.*|^6.0|^7.0|^8.0"
+        "illuminate/support": "5.*|^6.0|^7.0|^8.0",
+        "ext-json": "*"
     },
     "require-dev": {
         "orchestra/testbench": "^4.0|^5.0|^6.0",

config/config.php

@@ -82,6 +82,14 @@
              */
             'voice_id' => env('AWS_VOICE_ID', 'Amy'),
 
+            /**
+             * You can request any or all of the speech mark types, but leave it empty if you don't use speech marks.
+             * You may add any of the following:.
+             *
+             * sentence, word, viseme, ssml
+             */
+            'speech_marks' => [],
+
             /**
              * IAM Credentials from AWS.
              */

src/Converters/PollyConverter.php

@@ -6,14 +6,15 @@
 use Aws\Result;
 use Cion\TextToSpeech\Contracts\Converter;
 use Cion\TextToSpeech\Traits\HasLanguage;
+use Cion\TextToSpeech\Traits\HasSpeechMarks;
 use Cion\TextToSpeech\Traits\Sourceable;
 use Cion\TextToSpeech\Traits\SSMLable;
 use Cion\TextToSpeech\Traits\Storable;
 use Illuminate\Support\Arr;
 
 class PollyConverter implements Converter
 {
-    use Storable, Sourceable, HasLanguage, SSMLable;
+    use Storable, Sourceable, HasLanguage, SSMLable, HasSpeechMarks;
 
     /**
      * Client instance of Polly.
@@ -47,7 +48,7 @@ public function getClient(): PollyClient
      *
      * @param string $data
      * @param array $options
-     * @return string
+     * @return string|array
      */
     public function convert(string $data, array $options = null)
     {
@@ -59,6 +60,10 @@ public function convert(string $data, array $options = null)
 
         $result = $this->synthesizeSpeech($text, $options);
 
+        if ($this->hasSpeechMarks()) {
+            return $this->formatToArray($this->getResultContent($result));
+        }
+
         if ($result instanceof Result) {
             // Store audio file to disk
             return $this->store(
@@ -82,26 +87,28 @@ public function convert(string $data, array $options = null)
      */
     protected function synthesizeSpeech($text, array $options = null)
     {
+        $speechMarks = $this->getSpeechMarks();
+
+        $arguments = [
+            'LanguageCode'    => $this->getLanguage(),
+            'VoiceId'         => $this->voice($options),
+            'OutputFormat'    => $this->format($options),
+            'TextType'        => $this->textType(),
+            'SpeechMarkTypes' => $speechMarks,
+        ];
+
         if (is_string($text)) {
-            return $this->client->synthesizeSpeech([
-                'LanguageCode' => $this->getLanguage(),
-                'VoiceId'      => $this->voice($options),
-                'OutputFormat' => $this->format($options),
-                'TextType'     => $this->textType(),
-                'Text'         => $text,
-            ]);
+            return $this->client->synthesizeSpeech(array_merge($arguments, [
+                'Text' => $text,
+            ]));
         }
 
         $results = [];
 
         foreach ($text as $textItem) {
-            $result = $this->client->synthesizeSpeech([
-                'LanguageCode' => $this->getLanguage(),
-                'VoiceId'      => $this->voice($options),
-                'OutputFormat' => $this->format($options),
-                'TextType'     => $this->textType(),
-                'Text'         => $textItem,
-            ]);
+            $result = $this->client->synthesizeSpeech(array_merge($arguments, [
+                'Text' => $textItem,
+            ]));
 
             array_push($results, $result);
         }
@@ -190,11 +197,27 @@ protected function getLanguage()
      */
     protected function format($options)
     {
+        if ($this->hasSpeechMarks()) {
+            return 'json';
+        }
+
         $default = config('tts.output_format', 'mp3');
 
         return Arr::get($options, 'format', $default);
     }
 
+    /**
+     * Get the speech marks.
+     *
+     * @return array
+     */
+    protected function getSpeechMarks()
+    {
+        $default = config('tts.services.polly.speech_marks', []);
+
+        return ! empty($this->speechMarks) ? $this->speechMarks : $default;
+    }
+
     /**
      * Get the content of the result from AWS Polly.
      *
@@ -205,4 +228,31 @@ protected function getResultContent($result)
     {
         return $result->get('AudioStream')->getContents();
     }
+
+    /**
+     * Determines if speech marks are set.
+     *
+     * @return bool
+     */
+    protected function hasSpeechMarks()
+    {
+        return ! empty($this->getSpeechMarks());
+    }
+
+    /**
+     * Format the given json string into an array.
+     *
+     * @param string $json
+     * @return array
+     */
+    protected function formatToArray($json)
+    {
+        $jsons = explode(PHP_EOL, $json);
+
+        array_pop($jsons);
+
+        return collect($jsons)->map(function ($json) {
+            return json_decode($json, true);
+        })->toArray();
+    }
 }

src/Facades/TextToSpeech.php

@@ -5,12 +5,13 @@
 use Illuminate\Support\Facades\Facade;
 
 /**
- * @method static void convert($data, array $options)
+ * @method static string|array convert($data, array $options)
  * @method static \Cion\TextToSpeech\Contracts\Converter saveTo(string $path)
  * @method static \Cion\TextToSpeech\Contracts\Converter disk(string $disk)
  * @method static \Cion\TextToSpeech\Contracts\Converter source(string $source)
  * @method static \Cion\TextToSpeech\Contracts\Converter language(string $language)
  * @method static \Cion\TextToSpeech\Contracts\Converter ssml()
+ * @method static \Cion\TextToSpeech\Contracts\Converter speechMarks(array $speechMarks)
  */
 class TextToSpeech extends Facade
 {

src/Traits/HasSpeechMarks.php

@@ -0,0 +1,24 @@
+<?php
+
+namespace Cion\TextToSpeech\Traits;
+
+trait HasSpeechMarks
+{
+    /**
+     * @var array
+     */
+    protected $speechMarks = [];
+
+    /**
+     * Sets the speech mark types.
+     *
+     * @param array $speechMarks
+     * @return $this
+     */
+    public function speechMarks(array $speechMarks)
+    {
+        $this->speechMarks = $speechMarks;
+
+        return $this;
+    }
+}

tests/PollyConverterTest.php

@@ -97,4 +97,14 @@ public function it_should_specify_path_source_and_be_able_to_retrieve_text()
 
         Storage::assertExists($path);
     }
+
+    /** @test */
+    public function it_should_return_an_array_when_speech_marks_are_set()
+    {
+        $converter = new PollyConverter($this->client);
+
+        $result = $converter->speechMarks(['sentence'])->convert('test');
+
+        $this->assertIsArray($result);
+    }
 }