From 715f9ed46bbbff846689b212fdf78668971bcba8 Mon Sep 17 00:00:00 2001
From: kwaroran <kwaworan@gmail.com>
Date: Tue, 7 Nov 2023 05:02:34 +0900
Subject: [PATCH] [feat] oai tts

---
 src/lib/SideBars/CharConfig.svelte | 12 +++++++-
 src/ts/process/tts.ts              | 47 ++++++++++++++++++++++++++++++
 src/ts/storage/database.ts         |  1 +
 src/ts/translator/translator.ts    |  3 --
 4 files changed, 59 insertions(+), 4 deletions(-)

diff --git a/src/lib/SideBars/CharConfig.svelte b/src/lib/SideBars/CharConfig.svelte
index c1e91178..7ca95044 100644
--- a/src/lib/SideBars/CharConfig.svelte
+++ b/src/lib/SideBars/CharConfig.svelte
@@ -15,7 +15,7 @@
     import Help from "../Others/Help.svelte";
     import RegexData from "./Scripts/RegexData.svelte";
     import { exportChar, shareRisuHub } from "src/ts/characterCards";
-    import { getElevenTTSVoices, getWebSpeechTTSVoices, getVOICEVOXVoices } from "src/ts/process/tts";
+    import { getElevenTTSVoices, getWebSpeechTTSVoices, getVOICEVOXVoices, oaiVoices } from "src/ts/process/tts";
     import { checkCharOrder, getFileSrc } from "src/ts/storage/globalApi";
     import { addGroupChar, rmCharFromGroup } from "src/ts/process/group";
     import RealmUpload from "../UI/Realm/RealmUpload.svelte";
@@ -529,6 +529,7 @@
             <OptionInput value="elevenlab">ElevenLabs</OptionInput>
             <OptionInput value="webspeech">Web Speech</OptionInput>
             <OptionInput value="VOICEVOX">VOICEVOX</OptionInput>
+            <OptionInput value="openai">OpenAI</OptionInput>
         </SelectInput>
         
 
@@ -588,6 +589,15 @@
                 <NumberInput size={"sm"} marginBottom bind:value={currentChar.data.voicevoxConfig.INTONATION_SCALE}/>
                 <span class="text-sm mb-2 text-textcolor2">To use VOICEVOX, you need to run a colab and put the localtunnel URL in "Settings → Other Bots". https://colab.research.google.com/drive/1tyeXJSklNfjW-aZJAib1JfgOMFarAwze</span>
         {/if}
+        {#if currentChar.data.ttsMode === 'openai'}
+        <span class="text-textcolor">OpenAI TTS uses your OpenAI key on the chat model section</span>
+            <SelectInput className="mb-4 mt-2" bind:value={currentChar.data.oaiVoice}>
+                <OptionInput value="">Unset</OptionInput>
+                {#each oaiVoices as voice}
+                    <OptionInput value={voice}>{voice}</OptionInput>
+                {/each}
+            </SelectInput>
+        {/if}
         {#if currentChar.data.ttsMode === 'webspeech' || currentChar.data.ttsMode === 'elevenlab' || currentChar.data.ttsMode === 'VOICEVOX'}
             <div class="flex items-center mt-2">
                 <Check bind:check={currentChar.data.ttsReadOnlyQuoted} name={language.ttsReadOnlyQuoted}/>
diff --git a/src/ts/process/tts.ts b/src/ts/process/tts.ts
index 2c17908d..883e5b8f 100644
--- a/src/ts/process/tts.ts
+++ b/src/ts/process/tts.ts
@@ -2,6 +2,8 @@ import { get } from "svelte/store";
 import { alertError } from "../alert";
 import { DataBase, type character } from "../storage/database";
 import { translateVox } from "../translator/translator";
+import { globalFetch } from "../storage/globalApi";
+import { language } from "src/lang";
 
 let sourceNode:AudioBufferSourceNode = null
 
@@ -94,10 +96,55 @@ export async function sayTTS(character:character,text:string) {
                 }
             }
         }
+        case 'openai':{
+            const key = db.openAIKey
+            const res = await globalFetch('https://api.openai.com/v1/audio/speech', {
+                method: 'POST',
+                headers: {
+                    'Content-Type': 'application/json',
+                    'Authorization': 'Bearer ' + key,
+                },
+                body: {
+                    model: 'tts-1',
+                    input: text,
+                    voice: character.oaiVoice,
+                    
+                },
+                rawResponse: true,
+            })
+            const dat = res.data
+
+            if(res.ok){
+                try {
+                    const audio = Buffer.from(dat).buffer
+                    const audioContext = new AudioContext();
+                    const audioBuffer = await audioContext.decodeAudioData(audio)
+                    sourceNode = audioContext.createBufferSource();
+                    sourceNode.buffer = audioBuffer;
+                    sourceNode.connect(audioContext.destination);
+                    sourceNode.start();
+                } catch (error) {                    
+                    alertError(language.errors.httpError + `${error}`)
+                }
+            }
+            else{
+                if(dat.error && dat.error.message){                    
+                    alertError((language.errors.httpError + `${dat.error.message}`))
+                }
+                else{                    
+                    alertError((language.errors.httpError + `${JSON.stringify(res.data)}`))
+                }
+            }
+
+        }
     }
 
 }
 
+export const oaiVoices = [
+    'alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'
+]
+
 export function stopTTS(){
     if(sourceNode){
         sourceNode.stop()
diff --git a/src/ts/storage/database.ts b/src/ts/storage/database.ts
index 9d0e7825..455c3397 100644
--- a/src/ts/storage/database.ts
+++ b/src/ts/storage/database.ts
@@ -574,6 +574,7 @@ export interface character{
     license?:string
     private?:boolean
     additionalText:string
+    oaiVoice?:string
 }
 
 
diff --git a/src/ts/translator/translator.ts b/src/ts/translator/translator.ts
index 9e1b71b4..5c60bc45 100644
--- a/src/ts/translator/translator.ts
+++ b/src/ts/translator/translator.ts
@@ -155,8 +155,5 @@ export async function translateVox(text:string) {
 
 
 async function jaTrans(text:string) {
-    if(/[\u3000-\u303f\u3040-\u309f\u30a0-\u30ff\uff00-\uff9f\u4e00-\u9faf\u3400-\u4dbf]/.test(text)){
-        return text
-    }
     return await runTranslator(text,false, 'en','ja')
 }
\ No newline at end of file