[feat] hf tts translation

This commit is contained in:
kwaroran
2023-11-28 03:36:08 +09:00
parent a6b11c22de
commit 6ae8264f8b
7 changed files with 36 additions and 5 deletions

View File

@@ -1,5 +1,5 @@
<script lang="ts">
import { ArrowLeft, ArrowRight, PencilIcon, LanguagesIcon, RefreshCcwIcon, TrashIcon, CopyIcon } from "lucide-svelte";
import { ArrowLeft, ArrowRight, PencilIcon, LanguagesIcon, RefreshCcwIcon, TrashIcon, CopyIcon, Volume2Icon } from "lucide-svelte";
import { ParseMarkdown, type simpleCharacterArgument } from "../../ts/parser";
import AutoresizeArea from "../UI/GUI/TextAreaResizable.svelte";
import { alertConfirm, alertError } from "../../ts/alert";
@@ -10,6 +10,7 @@
import { risuChatParser } from "src/ts/process/scripts";
import { get } from "svelte/store";
import { isEqual } from "lodash";
import { sayTTS } from "src/ts/process/tts";
export let message = ''
export let name = ''
export let largePortrait = false
@@ -145,6 +146,12 @@
</button>
{/if}
{#if idx > -1}
<button class="ml-2 hover:text-green-500 transition-colors" on:click={()=>{
return sayTTS(null, message)
}}>
<Volume2Icon size={20}/>
</button>
<button class={"ml-2 hover:text-green-500 transition-colors "+(editMode?'text-green-400':'')} on:click={() => {
if(!editMode){
editMode = true

View File

@@ -301,7 +301,7 @@ export function characterFormatUpdate(index:number|character){
depth: 0,
prompt: ''
}
cha.hfTTS = {
cha.hfTTS ??= {
model: '',
language: 'en'
}

View File

@@ -1,5 +1,6 @@
import { get } from 'svelte/store'
import type { ScriptMode } from '../process/scripts'
//@ts-ignore
import WorkerUrl from './embedworker?worker&url'
import { DataBase, type Chat, type character, type Message } from '../storage/database'
import { selectedCharID } from '../stores'

View File

@@ -53,3 +53,10 @@ export const runEmbedding = async (text: string):Promise<Float32Array> => {
let result = await extractor(text, { pooling: 'mean', normalize: true });
return result?.data ?? null;
}
export const runTTS = async (text: string) => {
let speaker_embeddings = 'https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/speaker_embeddings.bin';
let synthesizer = await pipeline('text-to-speech', 'Xenova/speecht5_tts', { local_files_only: true });
let out = await synthesizer(text, { speaker_embeddings });
return out
}

View File

@@ -1,14 +1,21 @@
import { get } from "svelte/store";
import { alertError } from "../alert";
import { DataBase, type character } from "../storage/database";
import { translateVox } from "../translator/translator";
import { runTranslator, translateVox } from "../translator/translator";
import { globalFetch } from "../storage/globalApi";
import { language } from "src/lang";
import { sleep } from "../util";
import { getCurrentCharacter, sleep } from "../util";
let sourceNode:AudioBufferSourceNode = null
export async function sayTTS(character:character,text:string) {
if(!character){
const v = getCurrentCharacter()
if(v.type === 'group'){
return
}
character = v
}
let db = get(DataBase)
text = text.replace(/\*/g,'')
@@ -164,6 +171,9 @@ export async function sayTTS(character:character,text:string) {
}
case 'huggingface': {
while(true){
if(character.hfTTS.language !== 'en'){
text = await runTranslator(text, false, 'en', character.hfTTS.language)
}
const audioContext = new AudioContext();
const response = await fetch(`https://api-inference.huggingface.co/models/${character.hfTTS.model}`, {
method: 'POST',

View File

@@ -33,7 +33,7 @@ export async function translate(text:string, reverse:boolean) {
return runTranslator(text, reverse, db.translator,db.aiModel.startsWith('novellist') ? 'ja' : 'en')
}
async function runTranslator(text:string, reverse:boolean, from:string,target:'en'|'ja') {
export async function runTranslator(text:string, reverse:boolean, from:string,target:string) {
const arg = {
from: reverse ? from : target,

View File

@@ -365,4 +365,10 @@ export async function decryptBuffer(data:Uint8Array, keys:string){
)
return result
}
export function getCurrentCharacter(){
const db = get(DataBase)
const selectedChar = get(selectedCharID)
return db.characters[selectedChar]
}