WIP: Implement TTS coefont

discord/speeches/coefont.ts

@@ -0,0 +1,36 @@
+import axios from 'axios';
+import {SynthesizeFunction} from './types';
+
+const authorization = 'Bearer **********';
+let prevId: string = null;
+
+const speech: SynthesizeFunction = async (text: string, voiceType: string, {speed, engine}: {speed: number, engine: string}) => {
+	console.log('start', text);
+	const res = await axios.post('https://plbwpbyme3.execute-api.ap-northeast-1.amazonaws.com/production/projects/3d03e5ea-4596-43ce-a2c9-35450496d901/parts/e1106b18-03f0-4859-9cf5-e1cdc92a23ca/blocks', JSON.stringify({
+		coefontId: '19d55439-312d-4a1d-a27b-28f0f31bedc5',
+		text,
+		...(prevId ? {prevId} : {}),
+	}), {
+		headers: {
+			authorization,
+			'content-type': 'application/json',
+		},
+	});
+	console.log(res.data);
+	const {blockId} = res.data;
+	prevId = blockId;
+	const res2 = await axios.get(`https://plbwpbyme3.execute-api.ap-northeast-1.amazonaws.com/production/projects/3d03e5ea-4596-43ce-a2c9-35450496d901/parts/e1106b18-03f0-4859-9cf5-e1cdc92a23ca/blocks/${blockId}/audio`, {
+		headers: {
+			authorization,
+		},
+	});
+	console.log(res2.data);
+	const {location} = res2.data;
+	const res3 = await axios.get(location, {
+		responseType: 'arraybuffer',
+	});
+	console.log(res3.data);
+	return {data: res3.data as Buffer};
+};
+
+export default speech;

discord/speeches/index.ts

@@ -4,6 +4,7 @@ import azure from './azure';
 import google from './google';
 import voicetext, {Emotion, EmoLV} from './voicetext';
 import voicevox from './voicevox';
+import coefont from './coefont';
 
 const log = logger.child({bot: 'discord'});
 
@@ -45,6 +46,7 @@ enum Voice {
 	AI = 'AI',
 	AJ = 'AJ',
 	AK = 'AK',
+	AL = 'AL',
 }
 export {Voice};
 
@@ -61,7 +63,7 @@ export const getDefaultVoiceMeta: () => VoiceMeta = () => ({
 });
 
 interface Config {
-	provider: 'google' | 'amazon' | 'azure' | 'voicetext' | 'voicevox',
+	provider: 'google' | 'amazon' | 'azure' | 'voicetext' | 'voicevox' | 'coefont',
 	name: string,
 	emotional?: boolean,
 	lang?: string,
@@ -105,6 +107,7 @@ export const speechConfig: Map<Voice, Config> = new Map([
 	[Voice.AI, {provider: 'google', name: 'ja-JP-Neural2-B', lang: 'ja-JP'}],
 	[Voice.AJ, {provider: 'google', name: 'ja-JP-Neural2-C', lang: 'ja-JP'}],
 	[Voice.AK, {provider: 'google', name: 'ja-JP-Neural2-D', lang: 'ja-JP'}],
+	[Voice.AL, {provider: 'coefont', name: 'Hiroyuki', lang: 'ja-JP'}],
 	// coming soon
 	// [Voice., {provider: 'voicevox', name: 'whitecul', emotional: true}],
 	// [Voice., {provider: 'voicevox', name: 'goki', emotional: true}],
@@ -133,5 +136,8 @@ export const getSpeech = (text: string, voiceType: Voice, meta: VoiceMeta, audio
 	if (config.provider === 'voicevox') {
 		return voicevox(text, config.name, meta);
 	}
+	if (config.provider === 'coefont') {
+		return coefont(text, config.name, meta);
+	}
 	return voicetext(text, config.name, meta);
 };

discord/tts.ts

@@ -178,7 +178,7 @@ export default class TTS extends EventEmitter {
 		const tokens = message.content.split(/\s+/);
 		const user = message.member.user.id;
 
-		if (tokens[0]?.toUpperCase() === 'TTS') {
+		if (tokens[0]?.toUpperCase() === 'TTSDEV') {
 			if (tokens.length === 1 || tokens[1] === 'start') {
 				mutex.runExclusive(async () => {
 					if (!this.users.has(user)) {
@@ -246,6 +246,8 @@ export default class TTS extends EventEmitter {
 						providerName = 'VoiceText Web API';
 					} else if (config.provider === 'voicevox') {
 						providerName = 'VoiceVox Web API';
+					} else if (config.provider === 'coefont') {
+						providerName = 'CoeFont';
 					} else {
 						providerName = 'Unknown';
 					}