feat: add Microsoft Azure Speech to Text API

src/_locales/en/messages.json

@@ -30,7 +30,12 @@
   },
 
   "optionValue_speechService_ibmSpeechApi": {
-    "message": "IBM Speech to Text",
+    "message": "IBM Watson Speech to Text API",
+    "description": "Value of the option."
+  },
+
+  "optionValue_speechService_microsoftSpeechApi": {
+    "message": "Microsoft Azure Speech to Text API",
     "description": "Value of the option."
   },
 
@@ -64,6 +69,51 @@
     "description": "Value of the option."
   },
 
+  "optionTitle_microsoftSpeechApiLoc": {
+    "message": "API location",
+    "description": "Title of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_eastUs": {
+    "message": "East US",
+    "description": "Value of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_eastUs2": {
+    "message": "East US 2",
+    "description": "Value of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_westUs": {
+    "message": "West US",
+    "description": "Value of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_westUs2": {
+    "message": "West US 2",
+    "description": "Value of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_eastAsia": {
+    "message": "East Asia",
+    "description": "Value of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_southeastAsia": {
+    "message": "Southeast Asia",
+    "description": "Value of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_westEu": {
+    "message": "West Europe",
+    "description": "Value of the option."
+  },
+
+  "optionValue_microsoftSpeechApiLoc_northEu": {
+    "message": "North Europe",
+    "description": "Value of the option."
+  },
+
   "inputLabel_apiKey": {
     "message": "API key",
     "description": "Placeholder of the input."

src/options/App.vue

@@ -30,6 +30,19 @@
             :label="getText('inputLabel_apiKey')">
         </v-textfield>
       </div>
+      <div class="option select"
+          v-if="options.speechService === 'microsoftSpeechApi'">
+        <v-select :label="getText('optionTitle_microsoftSpeechApiLoc')"
+            v-model="options.microsoftSpeechApiLoc"
+            :options="selectOptions.microsoftSpeechApiLoc">
+        </v-select>
+      </div>
+      <div class="option text-field"
+          v-if="options.speechService === 'microsoftSpeechApi'">
+        <v-textfield v-model="options.microsoftSpeechApiKey"
+            :label="getText('inputLabel_apiKey')">
+        </v-textfield>
+      </div>
     </div>
   </div>
 </div>
@@ -58,22 +71,35 @@ export default {
         speechService: [
           'googleSpeechApiDemo',
           'googleSpeechApi',
-          'ibmSpeechApi'
+          'ibmSpeechApi',
+          'microsoftSpeechApi'
         ],
         ibmSpeechApiLoc: [
           'frankfurt',
           'dallas',
           'washington',
           'sydney',
           'tokyo'
+        ],
+        microsoftSpeechApiLoc: [
+          'eastUs',
+          'eastUs2',
+          'westUs',
+          'westUs2',
+          'eastAsia',
+          'southeastAsia',
+          'westEu',
+          'northEu'
         ]
       }),
 
       options: {
         speechService: '',
         googleSpeechApiKey: '',
         ibmSpeechApiLoc: '',
-        ibmSpeechApiKey: ''
+        ibmSpeechApiKey: '',
+        microsoftSpeechApiLoc: '',
+        microsoftSpeechApiKey: ''
       }
     };
   },

src/solve/main.js

@@ -6,7 +6,9 @@ import {getText, waitForElement, arrayBufferToBase64} from 'utils/common';
 import {
   captchaGoogleSpeechApiLangCodes,
   captchaIbmSpeechApiLangCodes,
-  ibmSpeechApiUrls
+  captchaMicrosoftSpeechApiLangCodes,
+  ibmSpeechApiUrls,
+  microsoftSpeechApiUrls
 } from 'utils/data';
 
 let solverWorking = false;
@@ -199,16 +201,56 @@ async function solve() {
       });
       return;
     }
+    const apiUrl = ibmSpeechApiUrls[apiLoc];
     const model = captchaIbmSpeechApiLangCodes[lang] || 'en-US_BroadbandModel';
 
+    const rsp = await fetch(`${apiUrl}?model=${model}&profanity_filter=false`, {
+      referrer: '',
+      mode: 'cors',
+      method: 'POST',
+      headers: {
+        Authorization: 'Basic ' + window.btoa('apiKey:' + apiKey)
+      },
+      body: new Blob([audioContent], {type: 'audio/wav'})
+    });
+
+    if (rsp.status !== 200) {
+      throw new Error(`API response: ${rsp.status}, ${await rsp.text()}`);
+    }
+
+    const results = (await rsp.json()).results;
+    if (results && results.length) {
+      solution = results[0].alternatives[0].transcript.trim();
+    }
+  }
+
+  if (speechService === 'microsoftSpeechApi') {
+    const {
+      microsoftSpeechApiLoc: apiLoc,
+      microsoftSpeechApiKey: apiKey
+    } = await storage.get(
+      ['microsoftSpeechApiLoc', 'microsoftSpeechApiKey'],
+      'sync'
+    );
+    if (!apiKey) {
+      browser.runtime.sendMessage({
+        id: 'notification',
+        messageId: 'error_missingApiKey'
+      });
+      return;
+    }
+    const apiUrl = microsoftSpeechApiUrls[apiLoc];
+    const language = captchaMicrosoftSpeechApiLangCodes[lang] || 'en-US';
+
     const rsp = await fetch(
-      `${ibmSpeechApiUrls[apiLoc]}?model=${model}&profanity_filter=false`,
+      `${apiUrl}?language=${language}&format=detailed&profanity=raw`,
       {
         referrer: '',
         mode: 'cors',
         method: 'POST',
         headers: {
-          Authorization: 'Basic ' + window.btoa('apiKey:' + apiKey)
+          'Ocp-Apim-Subscription-Key': apiKey,
+          'Content-type': 'audio/wav; codec=audio/pcm; samplerate=16000'
         },
         body: new Blob([audioContent], {type: 'audio/wav'})
       }
@@ -218,9 +260,9 @@ async function solve() {
       throw new Error(`API response: ${rsp.status}, ${await rsp.text()}`);
     }
 
-    const results = (await rsp.json()).results;
-    if (results && results.length) {
-      solution = results[0].alternatives[0].transcript.trim();
+    const results = (await rsp.json()).NBest;
+    if (results) {
+      solution = results[0].Lexical.trim();
     }
   }
 

src/storage/versions/local/ONiJBs00o.js

@@ -1,6 +1,6 @@
 import browser from 'webextension-polyfill';
 
-const message = 'Add IBM Speech to Text';
+const message = 'Add IBM Watson Speech to Text API';
 
 const revision = 'ONiJBs00o';
 const downRevision = 'UoT3kGyBH';
@@ -9,7 +9,7 @@ const storage = browser.storage.local;
 
 async function upgrade() {
   const changes = {
-    ibmSpeechApiLoc: 'frankfurt', // frankfurt, dallas, washington, sydney, tokyo
+    ibmSpeechApiLoc: 'frankfurt', // 'frankfurt', 'dallas', 'washington', 'sydney', 'tokyo'
     ibmSpeechApiKey: ''
   };
 

src/storage/versions/local/UidMDYaYA.js

@@ -0,0 +1,28 @@
+import browser from 'webextension-polyfill';
+
+const message = 'Add Microsoft Azure Speech to Text API';
+
+const revision = 'UidMDYaYA';
+const downRevision = 'ONiJBs00o';
+
+const storage = browser.storage.local;
+
+async function upgrade() {
+  const changes = {
+    microsoftSpeechApiLoc: 'eastUs', // 'eastUs', 'eastUs2', 'westUs', 'westUs2', 'eastAsia', 'southeastAsia', 'westEu', 'northEu'
+    microsoftSpeechApiKey: ''
+  };
+
+  changes.storageVersion = revision;
+  return storage.set(changes);
+}
+
+async function downgrade() {
+  const changes = {};
+  await storage.remove(['microsoftSpeechApiLoc', 'microsoftSpeechApiKey']);
+
+  changes.storageVersion = downRevision;
+  return storage.set(changes);
+}
+
+export {message, revision, upgrade, downgrade};

src/storage/versions/local/UoT3kGyBH.js

@@ -9,7 +9,7 @@ const storage = browser.storage.local;
 
 async function upgrade() {
   const changes = {
-    speechService: 'googleSpeechApiDemo', // 'googleSpeechApiDemo', 'googleSpeechApi'
+    speechService: 'googleSpeechApiDemo', // 'googleSpeechApiDemo', 'googleSpeechApi', 'ibmSpeechApi', 'microsoftSpeechApi'
     googleSpeechApiKey: '',
     installTime: new Date().getTime(),
     useCount: 0

src/storage/versions/local/versions.json

@@ -1 +1 @@
-{"versions": ["UoT3kGyBH", "ONiJBs00o"]}
+{"versions": ["UoT3kGyBH", "ONiJBs00o", "UidMDYaYA"]}

src/storage/versions/sync/ONiJBs00o.js

@@ -1,6 +1,6 @@
 import browser from 'webextension-polyfill';
 
-const message = 'Add IBM Speech to Text';
+const message = 'Add IBM Watson Speech to Text API';
 
 const revision = 'ONiJBs00o';
 const downRevision = 'UoT3kGyBH';
@@ -9,7 +9,7 @@ const storage = browser.storage.sync;
 
 async function upgrade() {
   const changes = {
-    ibmSpeechApiLoc: 'frankfurt', // frankfurt, dallas, washington, sydney, tokyo
+    ibmSpeechApiLoc: 'frankfurt', // 'frankfurt', 'dallas', 'washington', 'sydney', 'tokyo'
     ibmSpeechApiKey: ''
   };
 

src/storage/versions/sync/UidMDYaYA.js

@@ -0,0 +1,28 @@
+import browser from 'webextension-polyfill';
+
+const message = 'Add Microsoft Azure Speech to Text API';
+
+const revision = 'UidMDYaYA';
+const downRevision = 'ONiJBs00o';
+
+const storage = browser.storage.sync;
+
+async function upgrade() {
+  const changes = {
+    microsoftSpeechApiLoc: 'eastUs', // 'eastUs', 'eastUs2', 'westUs', 'westUs2', 'eastAsia', 'southeastAsia', 'westEu', 'northEu'
+    microsoftSpeechApiKey: ''
+  };
+
+  changes.storageVersion = revision;
+  return storage.set(changes);
+}
+
+async function downgrade() {
+  const changes = {};
+  await storage.remove(['microsoftSpeechApiLoc', 'microsoftSpeechApiKey']);
+
+  changes.storageVersion = downRevision;
+  return storage.set(changes);
+}
+
+export {message, revision, upgrade, downgrade};

src/storage/versions/sync/UoT3kGyBH.js

@@ -9,7 +9,7 @@ const storage = browser.storage.sync;
 
 async function upgrade() {
   const changes = {
-    speechService: 'googleSpeechApiDemo', // 'googleSpeechApiDemo', 'googleSpeechApi'
+    speechService: 'googleSpeechApiDemo', // 'googleSpeechApiDemo', 'googleSpeechApi', 'ibmSpeechApi', 'microsoftSpeechApi'
     googleSpeechApiKey: '',
     installTime: new Date().getTime(),
     useCount: 0

src/storage/versions/sync/versions.json

@@ -1 +1 @@
-{"versions": ["UoT3kGyBH", "ONiJBs00o"]}
+{"versions": ["UoT3kGyBH", "ONiJBs00o", "UidMDYaYA"]}