@ing.jorgeu/wargaming-overlay 2.4.0 → 3.0.0

package/tts/kokoro-assets.js

@@ -0,0 +1,27 @@
+const fs=require('node:fs/promises'),path=require('node:path'),os=require('node:os');
+const {createHash,randomUUID}=require('node:crypto');
+const MODEL='onnx-community/Kokoro-82M-v1.0-ONNX';
+const REVISION='1939ad2a8e416c0acfeecc08a694d14ef25f2231';
+const ASSETS={
+ 'onnx/model_quantized.onnx':'fbae9257e1e05ffc727e951ef9b9c98418e6d79f1c9b6b13bd59f5c9028a1478',
+ 'voices/ef_dora.bin':'f66ec66bd295acb18372e37008533a9a3228483ccd294e7538d5d9294ac9a532',
+ 'voices/em_alex.bin':'27809e9eafdcbcfff90a3016c697568676531de2a2c39cee29c96c7bd6b83e95',
+ 'voices/em_santa.bin':'ad43b774e1ca24d05c6161297d8aeb770ac3d29bb95daf516727af5f7d543683',
+ 'tokenizer.json':'77a02c8e164413299b4b4c403b14f8e0e1c1b727db4d46a09d6327b861060a34'
+};
+function cacheDirectory(){return process.env.KOKORO_CACHE_DIR||path.join(os.homedir(),'.cache','wargaming-overlay','kokoro',REVISION);}
+const hash=b=>createHash('sha256').update(b).digest('hex');
+async function asset(name,{cacheDir=cacheDirectory(),fetchImpl=fetch,signal,log=()=>{}}={}){
+ if(!Object.hasOwn(ASSETS,name))throw Error('Recurso Kokoro desconocido.');
+ const dest=path.join(cacheDir,name);
+ try{const b=await fs.readFile(dest);if(hash(b)===ASSETS[name])return b;}catch(e){if(e.code!=='ENOENT')throw e;}
+ log(`[Kokoro] Descargando ${name} (solo la primera vez).`);
+ const response=await fetchImpl(`https://huggingface.co/${MODEL}/resolve/${REVISION}/${name}`,{signal:signal||AbortSignal.timeout(90000)});
+ if(!response.ok)throw Error(`Descarga Kokoro: HTTP ${response.status}.`);
+ const data=Buffer.from(await response.arrayBuffer());if(hash(data)!==ASSETS[name])throw Error('La descarga Kokoro no pasó la verificación de integridad.');
+ await fs.mkdir(path.dirname(dest),{recursive:true});const temporary=dest+'.'+randomUUID()+'.tmp';
+ try{await fs.writeFile(temporary,data);await fs.rename(temporary,dest);}finally{await fs.rm(temporary,{force:true});}
+ return data;
+}
+async function ensureAssets(options){const [model,tokenizer,...voices]=await Promise.all(['onnx/model_quantized.onnx','tokenizer.json','voices/ef_dora.bin','voices/em_alex.bin','voices/em_santa.bin'].map(name=>asset(name,options)));return {model,voices:Object.fromEntries(['ef_dora','em_alex','em_santa'].map((v,i)=>[v,voices[i]])),vocab:JSON.parse(tokenizer).model.vocab};}
+module.exports={ensureAssets,asset,ASSETS,MODEL,REVISION,cacheDirectory};

package/tts/kokoro-engine.js

@@ -0,0 +1,31 @@
+const {ensureAssets}=require('./kokoro-assets');
+let engine;
+async function load(options={}){
+ if(!engine)engine=(async()=>{
+  const wasm=process.platform==='win32'||process.env.KOKORO_RUNTIME==='wasm';
+  const ort=require(wasm?'onnxruntime-web':'onnxruntime-node');
+  if(wasm)ort.env.wasm.numThreads=1;
+  const {model,voices,vocab}=await ensureAssets(options);
+  const {default:init}=await import('@echogarden/espeak-ng-emscripten');
+  const module=await init();const espeak=new module.eSpeakNGWorker();
+  const session=await ort.InferenceSession.create(model,{executionProviders:[wasm?'wasm':'cpu'],intraOpNumThreads:2});
+  return {ort,voices,vocab,espeak,session};
+ })().catch(e=>{engine=null;throw e;});
+ return engine;
+}
+async function generate(text,{voice='ef_dora',speed=1,...options}={}){
+ if(!['ef_dora','em_alex','em_santa'].includes(voice))throw Error('Voz española desconocida.');
+ if(!Number.isFinite(speed)||speed<0.5||speed>2)throw Error('kokoroSpeed debe estar entre 0.5 y 2.');
+ const {ort,voices,vocab,espeak,session}=await load(options);
+ espeak.set_voice('es');
+ // Preserve clause punctuation and map eSpeak IPA affricates to Kokoro's vocabulary.
+ const phones=String(text).normalize('NFC').split(/([:;,.!?])/).map(part=>/^[,:;.!?]$/.test(part)?part:espeak.synthesize_ipa(part.replace(/[<>\[\]]/g,'')).ipa.replaceAll('_','').trim().replace(/\n/g,' ')).join(' ').replaceAll('tʃ','ʧ').replaceAll('dʒ','ʤ').replace(/\u0361/g,'');
+ const tokens=Array.from(phones).filter(c=>vocab[c]!==undefined).map(c=>vocab[c]);
+ if(!tokens.length)throw Error('No hay texto pronunciable.');
+ if(tokens.length>509)throw Error('El texto excede el límite de Kokoro.');
+ const ids=[0,...tokens,0];const offset=tokens.length*256*4;
+ const bytes=voices[voice].subarray(offset,offset+1024);const style=new Float32Array(bytes.buffer.slice(bytes.byteOffset,bytes.byteOffset+bytes.length));
+ const out=await session.run({input_ids:new ort.Tensor('int64',BigInt64Array.from(ids,BigInt),[1,ids.length]),style:new ort.Tensor('float32',style,[1,256]),speed:new ort.Tensor('float32',Float32Array.of(speed),[1])});
+ return {samples:out.waveform.data,sampleRate:24000};
+}
+module.exports={generate};

package/tts/kokoro-worker.js

@@ -0,0 +1,13 @@
+// Isolate native synthesis/playback from the overlay. Killing this process cancels both.
+const {generate}=require('./kokoro-engine');
+const {play}=require('./playback');
+let input='';process.stdin.setEncoding('utf8');
+process.stdin.on('data',chunk=>{input+=chunk;if(input.length>20000)process.exit(1);});
+process.stdin.on('end',async()=>{
+ try{
+  const {text,voice,speed,cacheDir}=JSON.parse(input);const start=performance.now();
+  const {samples,sampleRate}=await generate(text,{voice,speed,cacheDir,log:message=>console.error(message)});
+  console.error(`[Kokoro] Audio generado en ${((performance.now()-start)/1000).toFixed(2)} s.`);
+  if(process.platform==='win32')await require('./windows-playback').playWindows(samples,sampleRate);else await play(samples,sampleRate);process.exit(0);
+ }catch(e){console.error('[Kokoro] '+e.message);process.exit(1);}
+});