microsoft sam tts

commands/interaction/tts/microsoft.js

@@ -0,0 +1,42 @@
+const { Constants } = require('detritus-client');
+const { InteractionCallbackTypes, ApplicationCommandOptionTypes } = Constants;
+
+const { sapi4 } = require('../../../labscore/api');
+const { MICROSOFT_VOICES, MICROSOFT_VOICE_CONFIG } = require('../../../labscore/constants');
+
+const { createEmbed } = require('../../../labscore/utils/embed');
+const { icon, highlight } = require('../../../labscore/utils/markdown');
+
+module.exports = {
+  description: 'Microsoft Sam Text to Speech',
+  name: 'mstts',
+  options: [
+    {
+      name: 'voice',
+      description: 'TTS Voice to use',
+      choices: MICROSOFT_VOICES,
+      required: true,
+    },
+    {
+      name: 'text',
+      description: 'Text',
+      type: ApplicationCommandOptionTypes.STRING,
+      required: true,
+      maxLength: 256
+    }
+  ],
+  run: async (context, args) => {
+    try{
+      await context.respond({data: {}, type: InteractionCallbackTypes.DEFERRED_CHANNEL_MESSAGE_WITH_SOURCE})
+      let audio = await sapi4(context, args.text, args.voice, MICROSOFT_VOICE_CONFIG[args.voice].pitch, MICROSOFT_VOICE_CONFIG[args.voice].speed)
+      await context.editOrRespond({
+        embeds: [createEmbed("defaultNoFooter", context, { description: `${icon("audio")} Audio Generated in ${highlight(audio.timings + "s")}.` })],
+        file: { value: audio.response.body, filename: "tts.wav" }
+      })
+    }catch(e){
+      await context.editOrRespond({
+        embeds: [createEmbed("error", context, "Unable to generate audio file.")]
+      })
+    }
+  },
+};

labscore/api/endpoints.js

@@ -46,6 +46,7 @@ const Api = Object.freeze({
   TTS_IMTRANSLATOR: '/tts/imtranslator',
   TTS_PLAYHT: '/tts/playht',
   TTS_POLLY: '/tts/polly',
+  TTS_SAPI4: '/tts/sapi4',
   TTS_TIKTOK: '/tts/tiktok',
   TTS_VOICEFORGE: '/tts/voiceforge',
 

labscore/api/index.js

@@ -268,6 +268,15 @@ module.exports.polly = async function(context, text, voice){
   })
 }
 
+module.exports.sapi4 = async function(context, text, voice, pitch = 50, speed = 150){
+  return await request(Api.TTS_SAPI4, "GET", {}, {
+    text,
+    voice,
+    pitch,
+    speed
+  })
+}
+
 module.exports.tiktok = async function(context, text, voice){
   return await request(Api.TTS_TIKTOK, "GET", {}, {
     text: text,

labscore/constants.js

@@ -278,6 +278,26 @@ module.exports.IMTRANSLATOR_VOICES = [
   { name: "Korean (Female)", value: "ko" }
 ]
 
+module.exports.MICROSOFT_VOICES = [
+  { name: "Microsoft Sam", value: "Sam" },
+  { name: "BonziBUDDY", value: "Bonzi" },
+  { name: "Mary", value: "Mary" },
+  { name: "Mike", value: "Mike" },
+  { name: "Robot", value: "RoboSoft Four" },
+  { name: "Whisper (Female)", value: "Female Whisper" },
+  { name: "Whisper (Male)", value: "Male Whisper" }
+]
+
+module.exports.MICROSOFT_VOICE_CONFIG = {
+  "Sam": { pitch: 100, speed: 150 },
+  "Bonzi": { pitch: 140, speed: 157 },
+  "Mary": { pitch: 169, speed: 170 },
+  "Mike": { pitch: 113, speed: 170 },
+  "RoboSoft Four": { pitch: 169, speed: 170 },
+  "Female Whisper": { pitch: 169, speed: 170 },
+  "Male Whisper": { pitch: 113, speed: 170 }
+}
+
 module.exports.TRANSLATE_LANGUAGE_MAPPINGS = Object.freeze({
   "af": "🇿🇦",
   "sq": "🇦🇱",