mincomk/risuai
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor: Update TTS function to handle auto audio path

Browse Source
This commit is contained in:
Junha Heo
2024-08-25 21:02:31 +09:00
parent 80bfc12461
commit bcb352942e
2 changed files with 74 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

74
src/lib/SideBars/CharConfig.svelte
View File

@@ -112,6 +112,13 @@
if (!(currentChar.data as character).gptSoVitsConfig.use_prompt) { if (!(currentChar.data as character).gptSoVitsConfig.use_prompt) {
(currentChar.data as character).gptSoVitsConfig.prompt = undefined (currentChar.data as character).gptSoVitsConfig.prompt = undefined
} }
if((currentChar.data as character).gptSoVitsConfig.use_auto_path){
(currentChar.data as character).gptSoVitsConfig.ref_audio_path = undefined;
(currentChar.data as character).gptSoVitsConfig.use_prompt = false;
(currentChar.data as character).gptSoVitsConfig.prompt = undefined;
}
} }
}) })
@@ -160,7 +167,9 @@
$: if (currentChar.data.ttsMode === 'gptsovits' && (currentChar.data as character).gptSoVitsConfig === undefined) { $: if (currentChar.data.ttsMode === 'gptsovits' && (currentChar.data as character).gptSoVitsConfig === undefined) {
(currentChar.data as character).gptSoVitsConfig = { (currentChar.data as character).gptSoVitsConfig = {
url: '', url: '',
ref_audio_path: 'C:/Users/user/Downloads/GPT-SoVITS-v2-240821', use_auto_path: false,
ref_audio_path: '',
use_long_audio: false,
ref_audio_data: { ref_audio_data: {
fileName: '', fileName: '',
assetId: '' assetId: ''
@@ -835,15 +844,24 @@
<span class="text-textcolor">URL</span> <span class="text-textcolor">URL</span>
<TextInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.url}/> <TextInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.url}/>
<span class="text-textcolor">Reference Audio Path (e.g. C:/Users/user/Downloads/GPT-SoVITS-v2-240821)</span> <span class="text-textcolor">Use Auto Path</span>
<TextInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.ref_audio_path}/> <Check bind:check={currentChar.data.gptSoVitsConfig.use_auto_path}/>
{#if !currentChar.data.gptSoVitsConfig.use_auto_path}
<span class="text-textcolor">Reference Audio Path (e.g. C:/Users/user/Downloads/GPT-SoVITS-v2-240821)</span>
<TextInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.ref_audio_path}/>
{/if}
<span class="text-textcolor">Use Long Audio</span>
<Check bind:check={currentChar.data.gptSoVitsConfig.use_long_audio}/>
<span class="text-textcolor">Reference Audio Data (3~10s audio file)</span> <span class="text-textcolor">Reference Audio Data (3~10s audio file)</span>
<Button on:click={async () => { <Button on:click={async () => {
const audio = await selectSingleFile([ const audio = await selectSingleFile([
'wav', 'wav',
'ogg', 'ogg',
'aac' 'aac',
'mp3'
]) ])
if(!audio){ if(!audio){
return null return null
@@ -866,40 +884,42 @@
</Button> </Button>
<span class="text-textcolor">Text Language</span> <span class="text-textcolor">Text Language</span>
<SelectInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.text_lang}> <SelectInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.text_lang}>
<OptionInput value="auto">Auto</OptionInput> <OptionInput value="auto">Multi-language Mixed</OptionInput>
<OptionInput value="auto_yue">Auto (Cantonese)</OptionInput> <OptionInput value="auto_yue">Multi-language Mixed (Cantonese)</OptionInput>
<OptionInput value="en">English</OptionInput> <OptionInput value="en">English</OptionInput>
<OptionInput value="zh">Chinese</OptionInput> <OptionInput value="zh">Chinese-English Mixed</OptionInput>
<OptionInput value="ja">Japanese</OptionInput> <OptionInput value="ja">Japanese-English Mixed</OptionInput>
<OptionInput value="yue">Cantonese</OptionInput> <OptionInput value="yue">Cantonese-English Mixed</OptionInput>
<OptionInput value="ko">Korean</OptionInput> <OptionInput value="ko">Korean-English Mixed</OptionInput>
<OptionInput value="all_zh">All Chinese</OptionInput> <OptionInput value="all_zh">Chinese</OptionInput>
<OptionInput value="all_ja">All Japanese</OptionInput> <OptionInput value="all_ja">Japanese</OptionInput>
<OptionInput value="all_yue">All Cantonese</OptionInput> <OptionInput value="all_yue">Cantonese</OptionInput>
<OptionInput value="all_ko">All Korean</OptionInput> <OptionInput value="all_ko">Korean</OptionInput>
</SelectInput> </SelectInput>
<span class="text-textcolor">Use Reference Audio Script</span> {#if !currentChar.data.gptSoVitsConfig.use_long_audio}
<Check bind:check={currentChar.data.gptSoVitsConfig.use_prompt}/> <span class="text-textcolor">Use Reference Audio Script</span>
<Check bind:check={currentChar.data.gptSoVitsConfig.use_prompt}/>
{/if}
{#if currentChar.data.gptSoVitsConfig.use_prompt} {#if currentChar.data.gptSoVitsConfig.use_prompt && !currentChar.data.gptSoVitsConfig.use_long_audio}
<span class="text-textcolor">Reference Audio Script</span> <span class="text-textcolor">Reference Audio Script</span>
<TextAreaInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.prompt}/> <TextAreaInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.prompt}/>
{/if} {/if}
<span class="text-textcolor">Reference Audio Language</span> <span class="text-textcolor">Reference Audio Language</span>
<SelectInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.prompt_lang}> <SelectInput className="mb-4 mt-2" bind:value={currentChar.data.gptSoVitsConfig.prompt_lang}>
<OptionInput value="auto">Auto</OptionInput> <OptionInput value="auto">Multi-language Mixed</OptionInput>
<OptionInput value="auto_yue">Auto (Cantonese)</OptionInput> <OptionInput value="auto_yue">Multi-language Mixed (Cantonese)</OptionInput>
<OptionInput value="en">English</OptionInput> <OptionInput value="en">English</OptionInput>
<OptionInput value="zh">Chinese</OptionInput> <OptionInput value="zh">Chinese-English Mixed</OptionInput>
<OptionInput value="ja">Japanese</OptionInput> <OptionInput value="ja">Japanese-English Mixed</OptionInput>
<OptionInput value="yue">Cantonese</OptionInput> <OptionInput value="yue">Cantonese-English Mixed</OptionInput>
<OptionInput value="ko">Korean</OptionInput> <OptionInput value="ko">Korean-English Mixed</OptionInput>
<OptionInput value="all_zh">English And Chinese</OptionInput> <OptionInput value="all_zh">Chinese</OptionInput>
<OptionInput value="all_ja">English And Japanese</OptionInput> <OptionInput value="all_ja">Japanese</OptionInput>
<OptionInput value="all_yue">English And Cantonese</OptionInput> <OptionInput value="all_yue">Cantonese</OptionInput>
<OptionInput value="all_ko">English And Korean</OptionInput> <OptionInput value="all_ko">Korean</OptionInput>
</SelectInput> </SelectInput>
<span class="text-textcolor">Top P</span> <span class="text-textcolor">Top P</span>
<SliderInput min={0.0} max={1.0} step={0.05} fixed={2} bind:value={currentChar.data.gptSoVitsConfig.top_p}/> <SliderInput min={0.0} max={1.0} step={0.05} fixed={2} bind:value={currentChar.data.gptSoVitsConfig.top_p}/>

31
src/ts/process/tts.ts
View File

@@ -235,12 +235,12 @@ export async function sayTTS(character:character,text:string) {
const audioContext = new AudioContext(); const audioContext = new AudioContext();
const audio: Uint8Array = await loadAsset(character.gptSoVitsConfig.ref_audio_data.assetId); const audio: Uint8Array = await loadAsset(character.gptSoVitsConfig.ref_audio_data.assetId);
const base64Audio = btoa(new Uint8Array(audio).reduce((data, byte) => data + String.fromCharCode(byte), '')); const base64Audio = btoa(new Uint8Array(audio).reduce((data, byte) => data + String.fromCharCode(byte), ''));
const body = { const body = {
text: text, text: text,
text_lang: character.gptSoVitsConfig.text_lang, text_lang: character.gptSoVitsConfig.text_lang,
ref_audio_path: character.gptSoVitsConfig.ref_audio_path + '/public/audio/' + character.gptSoVitsConfig.ref_audio_data.fileName, ref_audio_path: undefined,
ref_audio_name: character.gptSoVitsConfig.ref_audio_data.fileName, ref_audio_name: character.gptSoVitsConfig.ref_audio_data.fileName,
ref_audio_data: base64Audio, ref_audio_data: base64Audio,
prompt_text: undefined, prompt_text: undefined,
@@ -250,18 +250,41 @@ export async function sayTTS(character:character,text:string) {
speed_factor: character.gptSoVitsConfig.speed, speed_factor: character.gptSoVitsConfig.speed,
top_k: character.gptSoVitsConfig.top_k, top_k: character.gptSoVitsConfig.top_k,
text_split_method: character.gptSoVitsConfig.text_split_method, text_split_method: character.gptSoVitsConfig.text_split_method,
parallel_infer: false, parallel_infer: true,
// media_type: character.gptSoVitsConfig.ref_audio_data.fileName.split('.')[1],
ref_free: character.gptSoVitsConfig.use_long_audio || !character.gptSoVitsConfig.use_prompt,
} }
if (character.gptSoVitsConfig.use_prompt){ if (character.gptSoVitsConfig.use_prompt){
body.prompt_text = character.gptSoVitsConfig.prompt body.prompt_text = character.gptSoVitsConfig.prompt
} }
if (character.gptSoVitsConfig.use_auto_path){
console.log('auto')
const path = await globalFetch(`${character.gptSoVitsConfig.url}/get_path`, {
method: 'GET',
headers: {
'Content-Type': 'application/json'
},
rawResponse: false,
})
console.log(path)
if(path.ok){
body.ref_audio_path = path.data.message + '/public/audio/' + character.gptSoVitsConfig.ref_audio_data.fileName
}
else{
throw new Error('Failed to Auto get path')
}
} else {
body.ref_audio_path = character.gptSoVitsConfig.ref_audio_path + '/public/audio/' + character.gptSoVitsConfig.ref_audio_data.fileName
}
console.log(body) console.log(body)
const response = await globalFetch(`${character.gptSoVitsConfig.url}/tts`, { const response = await globalFetch(`${character.gptSoVitsConfig.url}/tts`, {
method: 'POST', method: 'POST',
headers: { headers: {
"Content-Type": "application/json", 'Content-Type': 'application/json'
}, },
body: body, body: body,
rawResponse: true, rawResponse: true,