Merge pull request #328 from harmony-one/voicehk

Added voicehk and voicehkf commands
harmony-one · Oct 2, 2023 · 79656ad · 79656ad
2 parents 41c8e38 + f24fbcb
commit 79656ad
Show file tree

Hide file tree

Showing 5 changed files with 72 additions and 10 deletions.
diff --git a/src/elevenlabs/elevenlabsClient.ts b/src/elevenlabs/elevenlabsClient.ts
@@ -0,0 +1,29 @@
+import axios, { type AxiosInstance } from 'axios'
+
+export class ElevenlabsClient {
+  private readonly _token: string
+  private readonly _httpClient: AxiosInstance
+  constructor (apiKey: string) {
+    this._token = apiKey
+
+    this._httpClient = axios.create({
+      baseURL: 'https://api.elevenlabs.io',
+      headers: {
+        'Content-Type': 'application/json',
+        Accept: 'audio/mpeg',
+        'xi-api-key': apiKey
+      }
+    })
+  }
+
+  public async textToSpeech ({ text, voiceId }: { text: string, voiceId: string }): Promise<string | Uint8Array | null | undefined> {
+    return await this._httpClient.post(`/v1/text-to-speech/${voiceId}`, {
+      text: 'string',
+      model_id: 'eleven_monolingual_v1',
+      voice_settings: {
+        stability: 0.5,
+        similarity_boost: 0.5
+      }
+    })
+  }
+}
diff --git a/src/elevenlabs/test.ts b/src/elevenlabs/test.ts
diff --git a/src/google-cloud/gcTextToSpeechClient.ts b/src/google-cloud/gcTextToSpeechClient.ts
@@ -2,18 +2,24 @@ import GcTextToSpeech, { type TextToSpeechClient } from '@google-cloud/text-to-s
 import config from '../config'
 import type { CredentialBody } from 'google-auth-library/build/src/auth/credentials'
 
+export interface TextToSpeechParams {
+  text: string
+  languageCode: 'en-US' | 'yue-Hant-HK' | 'ru-RU'
+  gender: 'MALE' | 'FEMALE'
+}
+
 class GcTextToSpeechClient {
   private readonly _client: TextToSpeechClient
   constructor (credentials: CredentialBody) {
     this._client = new GcTextToSpeech.TextToSpeechClient({ credentials })
   }
 
-  async textToSpeech (text: string): Promise<string | Uint8Array | null | undefined> {
+  async textToSpeech ({ text, languageCode, gender }: TextToSpeechParams): Promise<string | Uint8Array | null | undefined> {
     const ssml = `<speak>${text}</speak>`
 
     const [response] = await this._client.synthesizeSpeech({
       input: { ssml },
-      voice: { languageCode: 'en-US', ssmlGender: 'MALE' },
+      voice: { languageCode, ssmlGender: gender },
       audioConfig: { audioEncoding: 'OGG_OPUS' }
     })
 

diff --git a/src/modules/text-to-speech/index.ts b/src/modules/text-to-speech/index.ts
@@ -3,7 +3,14 @@ import { InputFile } from 'grammy'
 import type { Logger } from 'pino'
 import type { BotPayments } from '../payment'
 import type { OnMessageContext, PayableBot } from '../types'
-import { gcTextToSpeedClient } from '../../google-cloud/gcTextToSpeechClient'
+import { gcTextToSpeedClient, type TextToSpeechParams } from '../../google-cloud/gcTextToSpeechClient'
+
+enum SupportedCommands {
+  VOICE = 'voice',
+  VOICEHK = 'voicehk',
+  VOICEHKF = 'voicehkf',
+  VOICERU = 'voiceru',
+}
 
 export class TextToSpeechBot implements PayableBot {
   private readonly payments: BotPayments
@@ -22,7 +29,7 @@ export class TextToSpeechBot implements PayableBot {
   }
 
   public isSupportedEvent (ctx: OnMessageContext): boolean {
-    return ctx.hasCommand('voice')
+    return ctx.hasCommand(Object.values(SupportedCommands))
   }
 
   public getEstimatedPrice (ctx: OnMessageContext): number {
@@ -39,14 +46,34 @@ export class TextToSpeechBot implements PayableBot {
   }
 
   public async onEvent (ctx: OnMessageContext): Promise<void> {
-    if (ctx.hasCommand('voice')) {
+    if (ctx.hasCommand(SupportedCommands.VOICE)) {
+      const text = this.getTextFromMessage(ctx)
+      await this.onTextToSpeech(ctx, { text, gender: 'MALE', languageCode: 'en-US' })
+      return
+    }
+
+    if (ctx.hasCommand(SupportedCommands.VOICEHK)) {
       const text = this.getTextFromMessage(ctx)
-      await this.onTextToSpeech(ctx, text)
+      await this.onTextToSpeech(ctx, { text, gender: 'MALE', languageCode: 'yue-Hant-HK' })
+      return
+    }
+
+    if (ctx.hasCommand(SupportedCommands.VOICEHKF)) {
+      const text = this.getTextFromMessage(ctx)
+      await this.onTextToSpeech(ctx, { text, gender: 'FEMALE', languageCode: 'yue-Hant-HK' })
+      return
+    }
+
+    if (ctx.hasCommand(SupportedCommands.VOICERU)) {
+      const text = this.getTextFromMessage(ctx)
+      await this.onTextToSpeech(ctx, { text, gender: 'FEMALE', languageCode: 'ru-RU' })
     }
   }
 
-  public async onTextToSpeech (ctx: OnMessageContext, message: string): Promise<void> {
-    if (!message) {
+  public async onTextToSpeech (ctx: OnMessageContext, params: TextToSpeechParams): Promise<void> {
+    const { text, gender, languageCode } = params
+
+    if (!params.text) {
       await ctx.reply('/voice command should contain text.')
       return
     }
@@ -57,7 +84,7 @@ export class TextToSpeechBot implements PayableBot {
 
     const progressMessage = await ctx.reply('Generating...')
 
-    const voiceResult = await gcTextToSpeedClient.textToSpeech(message)
+    const voiceResult = await gcTextToSpeedClient.textToSpeech({ text, gender, languageCode })
 
     if (!voiceResult) {
       await ctx.api.editMessageText(ctx.chat.id, progressMessage.message_id, 'An error occurred during the process of generating the message.')

diff --git a/src/modules/voice-translate/index.ts b/src/modules/voice-translate/index.ts
@@ -71,7 +71,7 @@ export class VoiceTranslateBot implements PayableBot {
 
     const translateResult = await translator.translateText(resultText, null, 'en-US')
 
-    const voiceResult = await gcTextToSpeedClient.textToSpeech(translateResult.text)
+    const voiceResult = await gcTextToSpeedClient.textToSpeech({ text: translateResult.text, gender: 'MALE', languageCode: 'en-US' })
 
     if (!voiceResult) {
       await ctx.reply('voice generation error')