mellon 0.0.20 → 0.0.22

package/dist/mellon.cjs

@@ -1 +1 @@
-"use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});function Ut(m){return m&&m.__esModule&&Object.prototype.hasOwnProperty.call(m,"default")?m.default:m}var dt,pt;function Dt(){if(pt)return dt;pt=1;function m(o){if(this.size=o|0,this.size<=1||(this.size&this.size-1)!==0)throw new Error("FFT size must be a power of two and bigger than 1");this._csize=o<<1;for(var t=new Array(this.size*2),s=0;s<t.length;s+=2){const l=Math.PI*s/this.size;t[s]=Math.cos(l),t[s+1]=-Math.sin(l)}this.table=t;for(var n=0,e=1;this.size>e;e<<=1)n++;this._width=n%2===0?n-1:n,this._bitrev=new Array(1<<this._width);for(var c=0;c<this._bitrev.length;c++){this._bitrev[c]=0;for(var r=0;r<this._width;r+=2){var h=this._width-r-2;this._bitrev[c]|=(c>>>r&3)<<h}}this._out=null,this._data=null,this._inv=0}return dt=m,m.prototype.fromComplexArray=function(t,s){for(var n=s||new Array(t.length>>>1),e=0;e<t.length;e+=2)n[e>>>1]=t[e];return n},m.prototype.createComplexArray=function(){const t=new Array(this._csize);for(var s=0;s<t.length;s++)t[s]=0;return t},m.prototype.toComplexArray=function(t,s){for(var n=s||this.createComplexArray(),e=0;e<n.length;e+=2)n[e]=t[e>>>1],n[e+1]=0;return n},m.prototype.completeSpectrum=function(t){for(var s=this._csize,n=s>>>1,e=2;e<n;e+=2)t[s-e]=t[e],t[s-e+1]=-t[e+1]},m.prototype.transform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._transform4(),this._out=null,this._data=null},m.prototype.realTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._realTransform4(),this._out=null,this._data=null},m.prototype.inverseTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=1,this._transform4();for(var n=0;n<t.length;n++)t[n]/=this.size;this._out=null,this._data=null},m.prototype._transform4=function(){var t=this._out,s=this._csize,n=this._width,e=1<<n,c=s/e<<1,r,h,l=this._bitrev;if(c===4)for(r=0,h=0;r<s;r+=c,h++){const u=l[h];this._singleTransform2(r,u,e)}else for(r=0,h=0;r<s;r+=c,h++){const u=l[h];this._singleTransform4(r,u,e)}var i=this._inv?-1:1,a=this.table;for(e>>=2;e>=2;e>>=2){c=s/e<<1;var d=c>>>2;for(r=0;r<s;r+=c)for(var _=r+d,v=r,f=0;v<_;v+=2,f+=e){const u=v,p=u+d,w=p+d,g=w+d,T=t[u],A=t[u+1],b=t[p],y=t[p+1],F=t[w],S=t[w+1],C=t[g],E=t[g+1],M=T,P=A,R=a[f],U=i*a[f+1],D=b*R-y*U,z=b*U+y*R,k=a[2*f],N=i*a[2*f+1],B=F*k-S*N,j=F*N+S*k,K=a[3*f],O=i*a[3*f+1],$=C*K-E*O,G=C*O+E*K,J=M+B,x=P+j,I=M-B,Y=P-j,Q=D+$,L=z+G,W=i*(D-$),V=i*(z-G),tt=J+Q,rt=x+L,ot=J-Q,nt=x-L,at=I+V,it=Y-W,ct=I-V,lt=Y+W;t[u]=tt,t[u+1]=rt,t[p]=at,t[p+1]=it,t[w]=ot,t[w+1]=nt,t[g]=ct,t[g+1]=lt}}},m.prototype._singleTransform2=function(t,s,n){const e=this._out,c=this._data,r=c[s],h=c[s+1],l=c[s+n],i=c[s+n+1],a=r+l,d=h+i,_=r-l,v=h-i;e[t]=a,e[t+1]=d,e[t+2]=_,e[t+3]=v},m.prototype._singleTransform4=function(t,s,n){const e=this._out,c=this._data,r=this._inv?-1:1,h=n*2,l=n*3,i=c[s],a=c[s+1],d=c[s+n],_=c[s+n+1],v=c[s+h],f=c[s+h+1],u=c[s+l],p=c[s+l+1],w=i+v,g=a+f,T=i-v,A=a-f,b=d+u,y=_+p,F=r*(d-u),S=r*(_-p),C=w+b,E=g+y,M=T+S,P=A-F,R=w-b,U=g-y,D=T-S,z=A+F;e[t]=C,e[t+1]=E,e[t+2]=M,e[t+3]=P,e[t+4]=R,e[t+5]=U,e[t+6]=D,e[t+7]=z},m.prototype._realTransform4=function(){var t=this._out,s=this._csize,n=this._width,e=1<<n,c=s/e<<1,r,h,l=this._bitrev;if(c===4)for(r=0,h=0;r<s;r+=c,h++){const ht=l[h];this._singleRealTransform2(r,ht>>>1,e>>>1)}else for(r=0,h=0;r<s;r+=c,h++){const ht=l[h];this._singleRealTransform4(r,ht>>>1,e>>>1)}var i=this._inv?-1:1,a=this.table;for(e>>=2;e>=2;e>>=2){c=s/e<<1;var d=c>>>1,_=d>>>1,v=_>>>1;for(r=0;r<s;r+=c)for(var f=0,u=0;f<=v;f+=2,u+=e){var p=r+f,w=p+_,g=w+_,T=g+_,A=t[p],b=t[p+1],y=t[w],F=t[w+1],S=t[g],C=t[g+1],E=t[T],M=t[T+1],P=A,R=b,U=a[u],D=i*a[u+1],z=y*U-F*D,k=y*D+F*U,N=a[2*u],B=i*a[2*u+1],j=S*N-C*B,K=S*B+C*N,O=a[3*u],$=i*a[3*u+1],G=E*O-M*$,J=E*$+M*O,x=P+j,I=R+K,Y=P-j,Q=R-K,L=z+G,W=k+J,V=i*(z-G),tt=i*(k-J),rt=x+L,ot=I+W,nt=Y+tt,at=Q-V;if(t[p]=rt,t[p+1]=ot,t[w]=nt,t[w+1]=at,f===0){var it=x-L,ct=I-W;t[g]=it,t[g+1]=ct;continue}if(f!==v){var lt=Y,yt=-Q,Tt=x,bt=-I,At=-i*tt,Ft=-i*V,St=-i*W,Ct=-i*L,Et=lt+At,Mt=yt+Ft,Pt=Tt+Ct,Rt=bt-St,ut=r+_-f,vt=r+d-f;t[ut]=Et,t[ut+1]=Mt,t[vt]=Pt,t[vt+1]=Rt}}}},m.prototype._singleRealTransform2=function(t,s,n){const e=this._out,c=this._data,r=c[s],h=c[s+n],l=r+h,i=r-h;e[t]=l,e[t+1]=0,e[t+2]=i,e[t+3]=0},m.prototype._singleRealTransform4=function(t,s,n){const e=this._out,c=this._data,r=this._inv?-1:1,h=n*2,l=n*3,i=c[s],a=c[s+n],d=c[s+h],_=c[s+l],v=i+d,f=i-d,u=a+_,p=r*(a-_),w=v+u,g=f,T=-p,A=v-u,b=f,y=p;e[t]=w,e[t+1]=0,e[t+2]=g,e[t+3]=T,e[t+4]=A,e[t+5]=0,e[t+6]=b,e[t+7]=y},dt}var zt=Dt();const xt=Ut(zt);class ft{constructor(o=16e3,t=512,s=64){this._sampleRate=o,this._nfft=t,this._nfilt=s,this._fft=new xt(t),this._melFilters=this._createMelFilterbank()}_hzToMel(o){return 2595*Math.log10(1+o/700)}_melToHz(o){return 700*(10**(o/2595)-1)}_createMelFilterbank(){const t=this._sampleRate/2,s=this._hzToMel(0),n=this._hzToMel(t),e=new Float32Array(this._nfilt+2);for(let l=0;l<this._nfilt+2;l++)e[l]=s+l*(n-s)/(this._nfilt+1);const r=e.map(l=>this._melToHz(l)).map(l=>Math.floor((this._nfft+1)*l/this._sampleRate)),h=[];for(let l=0;l<this._nfilt;l++){const i=new Float32Array(Math.floor(this._nfft/2)+1);for(let a=r[l];a<r[l+1];a++)i[a]=(a-r[l])/(r[l+1]-r[l]);for(let a=r[l+1];a<r[l+2];a++)i[a]=(r[l+2]-a)/(r[l+2]-r[l+1]);h.push(i)}return h}logfbank(o){const t=Math.floor(.025*this._sampleRate),s=Math.floor(.01*this._sampleRate),n=1+Math.ceil((o.length-t)/s),e=new Float32Array(n*this._nfilt),c=new Float32Array(this._nfft),r=this._fft.createComplexArray();for(let h=0;h<n;h++){const l=h*s;c.fill(0);for(let d=0;d<t&&l+d<o.length;d++)c[d]=o[l+d];const i=this._fft.toComplexArray(c,null);this._fft.transform(r,i);const a=new Float32Array(Math.floor(this._nfft/2)+1);for(let d=0;d<a.length;d++){const _=r[2*d],v=r[2*d+1];a[d]=1/this._nfft*(_*_+v*v),a[d]===0&&(a[d]=1e-30)}for(let d=0;d<this._nfilt;d++){let _=0;const v=this._melFilters[d];for(let f=0;f<a.length;f++)_+=a[f]*v[f];_===0&&(_=1e-30),e[h*this._nfilt+d]=Math.log(_)}}return e}maxCosineSim(o,t){let s=0;for(const n of t){let e=0;for(let r=0;r<n.length;r++)e+=o[r]*n[r];const c=(e+1)/2;c>s&&(s=c)}return s}}async function st(m=Z,o=q){const t=await import(o);return t.env.wasm.wasmPaths=m,t.env.wasm.numThreads=1,t}let _t=null;async function mt(m=et,o=Z,t=q,s){return _t||(_t=st(o,t).then(n=>s?n.InferenceSession.create(new Uint8Array(s),{executionProviders:["wasm"],graphOptimizationLevel:"all"}):n.InferenceSession.create(m,{executionProviders:["wasm"],graphOptimizationLevel:"all"}))),_t}class H{static loadWords(o=X){try{const t=localStorage.getItem(o);return t?JSON.parse(t):[]}catch{return[]}}static saveWord(o,t=X){const s=H.loadWords(t).filter(n=>n.word_name!==o.word_name);localStorage.setItem(t,JSON.stringify([...s,o]))}static deleteWord(o,t=X){try{const s=H.loadWords(t).filter(n=>n.word_name!==o);localStorage.setItem(t,JSON.stringify(s))}catch{}}}class It{constructor(o,t){this._started=!1,this._inferring=!1,this._audioCtx=null,this._stream=null,this._refEmbeddings=new Map,this._lastMatchAt=0,this._lastInferenceAt=0,this._initPromise=null;const{refsStorageKey:s=X,thresholdStorageKey:n=gt,wasmPaths:e=Z,modelPath:c=et,audioProcessorPath:r=wt,ortCdnUrl:h=q,audioUtils:l=new ft}=t||{};this._audioUtils=l,this._commands=o,this._refsStorageKey=s,this._thresholdStorageKey=n,this._audioProcessorPath=r,this._wasmPaths=e,this._modelPath=c,this._ortCdnUrl=h;try{const i=localStorage.getItem(this._thresholdStorageKey);this._threshold=i!==null?Math.max(0,Math.min(1,Number(i))):.65}catch{this._threshold=.65}}get threshold(){return this._threshold}set threshold(o){this._threshold=Math.max(0,Math.min(1,o));try{localStorage.setItem(this._thresholdStorageKey,String(this._threshold))}catch{}}get listening(){return this._started}async _trackFetch(o,t,s){const n=await fetch(o);if(!n.ok)throw new Error(`HTTP ${n.status} fetching ${o}`);const e=Number(n.headers.get("content-length")??"0");if(e>0&&(s.total+=e),!n.body){const a=await n.arrayBuffer();return s.downloaded+=a.byteLength,e||(s.total+=a.byteLength),t==null||t(s.downloaded,s.total),a}const c=n.body.getReader(),r=[];let h=0;for(;;){const{done:a,value:d}=await c.read();if(a)break;r.push(d),h+=d.length,s.downloaded+=d.length,t==null||t(s.downloaded,s.total)}e||(s.total+=h);const l=new Uint8Array(h);let i=0;for(const a of r)l.set(a,i),i+=a.length;return l.buffer}async _init(o){const t={downloaded:0,total:0},s=new Set,n=[];for(const i of this._commands)for(const a of i.triggers)!s.has(a.name)&&a.defaultRefPath&&(s.add(a.name),n.push({name:a.name,path:a.defaultRefPath}));const e=st(this._wasmPaths,this._ortCdnUrl),[c,...r]=await Promise.all([this._trackFetch(this._modelPath,o,t),...n.map(({path:i})=>this._trackFetch(i,o,t))]);await e,await mt(this._modelPath,this._wasmPaths,this._ortCdnUrl,c);const h=H.loadWords(this._refsStorageKey),l=new Set(h.map(i=>i.word_name));for(let i=0;i<n.length;i++)try{const a=JSON.parse(new TextDecoder().decode(r[i]));this.addCustomWord(a),l.has(a.word_name)||H.saveWord(a,this._refsStorageKey)}catch{console.warn(`[Mellon] failed to parse ref file: ${n[i].path}`)}for(const i of h)this._refEmbeddings.set(i.word_name,i.embeddings);console.info("[Mellon] init complete, loaded refs:",[...this._refEmbeddings.keys()])}async init(o){this._initPromise||(this._initPromise=this._init(o)),await this._initPromise}addCustomWord(o){if(!(Array.isArray(o.embeddings)&&o.embeddings.length>0))throw new Error("invalid ref file for : "+o.word_name);this._refEmbeddings.set(o.word_name,o.embeddings)}async start(){if(this._started)return;await this.init();let o;try{o=await navigator.mediaDevices.getUserMedia({audio:{noiseSuppression:!1,echoCancellation:!1,autoGainControl:!1,channelCount:1}})}catch{o=await navigator.mediaDevices.getUserMedia({audio:!0})}this._stream=o;const t=new AudioContext({sampleRate:16e3});this._audioCtx=t,await t.audioWorklet.addModule(this._audioProcessorPath);const s=t.createMediaStreamSource(o),n=new AudioWorkletNode(t,"audio-processor");n.port.onmessage=e=>{this._handleBuffer(e.data)},s.connect(n),n.connect(t.destination),this._started=!0}async stop(){if(this._started=!1,this._audioCtx&&(await this._audioCtx.close(),this._audioCtx=null),this._stream){for(const o of this._stream.getTracks())o.stop();this._stream=null}}async _handleBuffer(o){if(this._inferring)return;const t=Date.now();if(!(t-this._lastInferenceAt<300)){this._lastInferenceAt=t,this._inferring=!0;try{const[s,n]=await Promise.all([st(this._wasmPaths,this._ortCdnUrl),mt(this._modelPath,this._wasmPaths,this._ortCdnUrl)]),e=this._audioUtils.logfbank(o),c=new s.Tensor("float32",e,[1,1,149,64]),r=await n.run({input:c}),h=r[Object.keys(r)[0]].data;let l=!1;for(const i of this._commands){if(l)break;for(const a of i.triggers){const d=this._refEmbeddings.get(a.name);if(!d)continue;const _=this._audioUtils.maxCosineSim(h,d);if(_>=this._threshold&&t-this._lastMatchAt>2e3){this._lastMatchAt=t,console.info(`[Mellon] match: "${a}" sim=${_.toFixed(3)}`),typeof i.onMatch=="function"&&i.onMatch(a.name,_),l=!0;break}}}}catch(s){console.error("[Mellon] inference error:",s)}finally{this._inferring=!1}}}}class Lt{constructor(o,t){this._config={},this._samples=[],this._wordName=o,this._config.modelPath=(t==null?void 0:t.modelPath)||et,this._config.wasmPaths=(t==null?void 0:t.wasmPaths)||Z,this._config.ortCdnUrl=(t==null?void 0:t.ortCdnUrl)||q,this._audioUtils=(t==null?void 0:t.audioUtils)??new ft}async recordSample(){const o=await navigator.mediaDevices.getUserMedia({audio:!0}),t=new AudioContext({sampleRate:16e3}),s=await new Promise((c,r)=>{const h=new MediaRecorder(o),l=[];h.ondataavailable=i=>{i.data.size>0&&l.push(i.data)},h.onstop=async()=>{var i;for(const a of o.getTracks())a.stop();try{const d=await new Blob(l,{type:((i=l[0])==null?void 0:i.type)||"audio/webm"}).arrayBuffer(),_=await t.decodeAudioData(d);await t.close(),c(_.getChannelData(0).slice())}catch(a){r(a)}},h.start(),setTimeout(()=>{try{h.stop()}catch{}},1500)}),n=24e3,e=new Float32Array(n);return e.set(s.slice(0,n)),this._samples.push(e),this._samples.length}deleteSample(o){if(o<0||o>=this._samples.length)throw new RangeError(`index ${o} out of bounds (${this._samples.length} samples)`);return this._samples.splice(o,1),this._samples.length}async generateRef(){const[o,t]=await Promise.all([st(this._config.wasmPaths,this._config.ortCdnUrl),mt(this._config.modelPath,this._config.wasmPaths,this._config.ortCdnUrl)]),s=[];for(const n of this._samples){const e=this._audioUtils.logfbank(n),c=new o.Tensor("float32",e,[1,1,149,64]),r=await t.run({input:c}),h=Array.from(r[Object.keys(r)[0]].data);s.push(h)}return{word_name:this._wordName,model_type:"resnet_50_arc",embeddings:s}}}const Z="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/",q="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs",et="https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx",wt="https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js",X="mellon-refs",gt="mellon-threshold";exports.AudioUtils=ft;exports.DEFAULT_AUDIO_PROCESSOR_PATH=wt;exports.DEFAULT_MODEL_PATH=et;exports.DEFAULT_ORT_CDN_URL=q;exports.DEFAULT_REFS_STORAGE_KEY=X;exports.DEFAULT_THRESHOLD_STORAGE_KEY=gt;exports.DEFAULT_WASM_PATHS=Z;exports.Detector=It;exports.EnrollmentSession=Lt;exports.Storage=H;
+"use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});function Ut(m){return m&&m.__esModule&&Object.prototype.hasOwnProperty.call(m,"default")?m.default:m}var _t,pt;function Dt(){if(pt)return _t;pt=1;function m(o){if(this.size=o|0,this.size<=1||(this.size&this.size-1)!==0)throw new Error("FFT size must be a power of two and bigger than 1");this._csize=o<<1;for(var t=new Array(this.size*2),s=0;s<t.length;s+=2){const c=Math.PI*s/this.size;t[s]=Math.cos(c),t[s+1]=-Math.sin(c)}this.table=t;for(var n=0,e=1;this.size>e;e<<=1)n++;this._width=n%2===0?n-1:n,this._bitrev=new Array(1<<this._width);for(var i=0;i<this._bitrev.length;i++){this._bitrev[i]=0;for(var r=0;r<this._width;r+=2){var h=this._width-r-2;this._bitrev[i]|=(i>>>r&3)<<h}}this._out=null,this._data=null,this._inv=0}return _t=m,m.prototype.fromComplexArray=function(t,s){for(var n=s||new Array(t.length>>>1),e=0;e<t.length;e+=2)n[e>>>1]=t[e];return n},m.prototype.createComplexArray=function(){const t=new Array(this._csize);for(var s=0;s<t.length;s++)t[s]=0;return t},m.prototype.toComplexArray=function(t,s){for(var n=s||this.createComplexArray(),e=0;e<n.length;e+=2)n[e]=t[e>>>1],n[e+1]=0;return n},m.prototype.completeSpectrum=function(t){for(var s=this._csize,n=s>>>1,e=2;e<n;e+=2)t[s-e]=t[e],t[s-e+1]=-t[e+1]},m.prototype.transform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._transform4(),this._out=null,this._data=null},m.prototype.realTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._realTransform4(),this._out=null,this._data=null},m.prototype.inverseTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=1,this._transform4();for(var n=0;n<t.length;n++)t[n]/=this.size;this._out=null,this._data=null},m.prototype._transform4=function(){var t=this._out,s=this._csize,n=this._width,e=1<<n,i=s/e<<1,r,h,c=this._bitrev;if(i===4)for(r=0,h=0;r<s;r+=i,h++){const u=c[h];this._singleTransform2(r,u,e)}else for(r=0,h=0;r<s;r+=i,h++){const u=c[h];this._singleTransform4(r,u,e)}var d=this._inv?-1:1,a=this.table;for(e>>=2;e>=2;e>>=2){i=s/e<<1;var l=i>>>2;for(r=0;r<s;r+=i)for(var _=r+l,v=r,f=0;v<_;v+=2,f+=e){const u=v,p=u+l,w=p+l,g=w+l,T=t[u],b=t[u+1],A=t[p],y=t[p+1],F=t[w],S=t[w+1],C=t[g],R=t[g+1],P=T,E=b,M=a[f],U=d*a[f+1],D=A*M-y*U,z=A*U+y*M,k=a[2*f],N=d*a[2*f+1],B=F*k-S*N,j=F*N+S*k,K=a[3*f],O=d*a[3*f+1],$=C*K-R*O,G=C*O+R*K,J=P+B,x=E+j,I=P-B,Y=E-j,Q=D+$,L=z+G,W=d*(D-$),V=d*(z-G),tt=J+Q,ot=x+L,nt=J-Q,at=x-L,it=I+V,ct=Y-W,lt=I-V,ht=Y+W;t[u]=tt,t[u+1]=ot,t[p]=it,t[p+1]=ct,t[w]=nt,t[w+1]=at,t[g]=lt,t[g+1]=ht}}},m.prototype._singleTransform2=function(t,s,n){const e=this._out,i=this._data,r=i[s],h=i[s+1],c=i[s+n],d=i[s+n+1],a=r+c,l=h+d,_=r-c,v=h-d;e[t]=a,e[t+1]=l,e[t+2]=_,e[t+3]=v},m.prototype._singleTransform4=function(t,s,n){const e=this._out,i=this._data,r=this._inv?-1:1,h=n*2,c=n*3,d=i[s],a=i[s+1],l=i[s+n],_=i[s+n+1],v=i[s+h],f=i[s+h+1],u=i[s+c],p=i[s+c+1],w=d+v,g=a+f,T=d-v,b=a-f,A=l+u,y=_+p,F=r*(l-u),S=r*(_-p),C=w+A,R=g+y,P=T+S,E=b-F,M=w-A,U=g-y,D=T-S,z=b+F;e[t]=C,e[t+1]=R,e[t+2]=P,e[t+3]=E,e[t+4]=M,e[t+5]=U,e[t+6]=D,e[t+7]=z},m.prototype._realTransform4=function(){var t=this._out,s=this._csize,n=this._width,e=1<<n,i=s/e<<1,r,h,c=this._bitrev;if(i===4)for(r=0,h=0;r<s;r+=i,h++){const dt=c[h];this._singleRealTransform2(r,dt>>>1,e>>>1)}else for(r=0,h=0;r<s;r+=i,h++){const dt=c[h];this._singleRealTransform4(r,dt>>>1,e>>>1)}var d=this._inv?-1:1,a=this.table;for(e>>=2;e>=2;e>>=2){i=s/e<<1;var l=i>>>1,_=l>>>1,v=_>>>1;for(r=0;r<s;r+=i)for(var f=0,u=0;f<=v;f+=2,u+=e){var p=r+f,w=p+_,g=w+_,T=g+_,b=t[p],A=t[p+1],y=t[w],F=t[w+1],S=t[g],C=t[g+1],R=t[T],P=t[T+1],E=b,M=A,U=a[u],D=d*a[u+1],z=y*U-F*D,k=y*D+F*U,N=a[2*u],B=d*a[2*u+1],j=S*N-C*B,K=S*B+C*N,O=a[3*u],$=d*a[3*u+1],G=R*O-P*$,J=R*$+P*O,x=E+j,I=M+K,Y=E-j,Q=M-K,L=z+G,W=k+J,V=d*(z-G),tt=d*(k-J),ot=x+L,nt=I+W,at=Y+tt,it=Q-V;if(t[p]=ot,t[p+1]=nt,t[w]=at,t[w+1]=it,f===0){var ct=x-L,lt=I-W;t[g]=ct,t[g+1]=lt;continue}if(f!==v){var ht=Y,yt=-Q,Tt=x,At=-I,bt=-d*tt,Ft=-d*V,St=-d*W,Ct=-d*L,Rt=ht+bt,Pt=yt+Ft,Et=Tt+Ct,Mt=At-St,ut=r+_-f,vt=r+l-f;t[ut]=Rt,t[ut+1]=Pt,t[vt]=Et,t[vt+1]=Mt}}}},m.prototype._singleRealTransform2=function(t,s,n){const e=this._out,i=this._data,r=i[s],h=i[s+n],c=r+h,d=r-h;e[t]=c,e[t+1]=0,e[t+2]=d,e[t+3]=0},m.prototype._singleRealTransform4=function(t,s,n){const e=this._out,i=this._data,r=this._inv?-1:1,h=n*2,c=n*3,d=i[s],a=i[s+n],l=i[s+h],_=i[s+c],v=d+l,f=d-l,u=a+_,p=r*(a-_),w=v+u,g=f,T=-p,b=v-u,A=f,y=p;e[t]=w,e[t+1]=0,e[t+2]=g,e[t+3]=T,e[t+4]=b,e[t+5]=0,e[t+6]=A,e[t+7]=y},_t}var zt=Dt();const xt=Ut(zt);class ft{constructor(o=16e3,t=512,s=64){this._sampleRate=o,this._nfft=t,this._nfilt=s,this._fft=new xt(t),this._melFilters=this._createMelFilterbank()}_hzToMel(o){return 2595*Math.log10(1+o/700)}_melToHz(o){return 700*(10**(o/2595)-1)}_createMelFilterbank(){const t=this._sampleRate/2,s=this._hzToMel(0),n=this._hzToMel(t),e=new Float32Array(this._nfilt+2);for(let c=0;c<this._nfilt+2;c++)e[c]=s+c*(n-s)/(this._nfilt+1);const r=e.map(c=>this._melToHz(c)).map(c=>Math.floor((this._nfft+1)*c/this._sampleRate)),h=[];for(let c=0;c<this._nfilt;c++){const d=new Float32Array(Math.floor(this._nfft/2)+1);for(let a=r[c];a<r[c+1];a++)d[a]=(a-r[c])/(r[c+1]-r[c]);for(let a=r[c+1];a<r[c+2];a++)d[a]=(r[c+2]-a)/(r[c+2]-r[c+1]);h.push(d)}return h}logfbank(o){const t=Math.floor(.025*this._sampleRate),s=Math.floor(.01*this._sampleRate),n=1+Math.ceil((o.length-t)/s),e=new Float32Array(n*this._nfilt),i=new Float32Array(this._nfft),r=this._fft.createComplexArray();for(let h=0;h<n;h++){const c=h*s;i.fill(0);for(let l=0;l<t&&c+l<o.length;l++)i[l]=o[c+l];const d=this._fft.toComplexArray(i,null);this._fft.transform(r,d);const a=new Float32Array(Math.floor(this._nfft/2)+1);for(let l=0;l<a.length;l++){const _=r[2*l],v=r[2*l+1];a[l]=1/this._nfft*(_*_+v*v),a[l]===0&&(a[l]=1e-30)}for(let l=0;l<this._nfilt;l++){let _=0;const v=this._melFilters[l];for(let f=0;f<a.length;f++)_+=a[f]*v[f];_===0&&(_=1e-30),e[h*this._nfilt+l]=Math.log(_)}}return e}maxCosineSim(o,t){let s=0;for(const n of t){let e=0;for(let r=0;r<n.length;r++)e+=o[r]*n[r];const i=(e+1)/2;i>s&&(s=i)}return s}}async function et(m=Z,o=q){const t=await import(o);return t.env.wasm.wasmPaths=m,t.env.wasm.numThreads=1,t}let st=null;async function mt(m=rt,o=Z,t=q,s){return st||(st=et(o,t).then(n=>s?n.InferenceSession.create(new Uint8Array(s),{executionProviders:["wasm"],graphOptimizationLevel:"all"}):n.InferenceSession.create(m,{executionProviders:["wasm"],graphOptimizationLevel:"all"}))),st}function It(){return st!==null}class H{static loadWords(o=X){try{const t=localStorage.getItem(o);return t?JSON.parse(t):[]}catch{return[]}}static saveWord(o,t=X){const s=H.loadWords(t).filter(n=>n.word_name!==o.word_name);localStorage.setItem(t,JSON.stringify([...s,o]))}static deleteWord(o,t=X){try{const s=H.loadWords(t).filter(n=>n.word_name!==o);localStorage.setItem(t,JSON.stringify(s))}catch{}}}class Lt{constructor(o,t){this._started=!1,this._inferring=!1,this._audioCtx=null,this._stream=null,this._refEmbeddings=new Map,this._lastMatchAt=0,this._lastInferenceAt=0,this._initPromise=null;const{refsStorageKey:s=X,thresholdStorageKey:n=gt,wasmPaths:e=Z,modelPath:i=rt,audioProcessorPath:r=wt,ortCdnUrl:h=q,audioUtils:c=new ft}=t||{};this._audioUtils=c,this._commands=o,this._refsStorageKey=s,this._thresholdStorageKey=n,this._audioProcessorPath=r,this._wasmPaths=e,this._modelPath=i,this._ortCdnUrl=h;try{const d=localStorage.getItem(this._thresholdStorageKey);this._threshold=d!==null?Math.max(0,Math.min(1,Number(d))):.65}catch{this._threshold=.65}}get threshold(){return this._threshold}set threshold(o){this._threshold=Math.max(0,Math.min(1,o));try{localStorage.setItem(this._thresholdStorageKey,String(this._threshold))}catch{}}get listening(){return this._started}async _trackFetch(o,t,s){const n=await fetch(o);if(!n.ok)throw new Error(`HTTP ${n.status} fetching ${o}`);const e=Number(n.headers.get("content-length")??"0");if(e>0&&(s.total+=e),!n.body){const a=await n.arrayBuffer();return s.downloaded+=a.byteLength,e||(s.total+=a.byteLength),t==null||t(s.downloaded,s.total),a}const i=n.body.getReader(),r=[];let h=0;for(;;){const{done:a,value:l}=await i.read();if(a)break;r.push(l),h+=l.length,s.downloaded+=l.length,t==null||t(s.downloaded,s.total)}e||(s.total+=h);const c=new Uint8Array(h);let d=0;for(const a of r)c.set(a,d),d+=a.length;return c.buffer}async _init(o){const t={downloaded:0,total:0},s=H.loadWords(this._refsStorageKey),n=new Set(s.map(a=>a.word_name)),e=new Set,i=[];for(const a of this._commands)for(const l of a.triggers)!e.has(l.name)&&l.defaultRefPath&&!n.has(l.name)&&(e.add(l.name),i.push({name:l.name,path:l.defaultRefPath}));const r=et(this._wasmPaths,this._ortCdnUrl),h=It(),[c,...d]=await Promise.all([h?Promise.resolve(null):this._trackFetch(this._modelPath,o,t),...i.map(({path:a})=>this._trackFetch(a,o,t))]);await r,await mt(this._modelPath,this._wasmPaths,this._ortCdnUrl,h?void 0:c);for(let a=0;a<i.length;a++)try{const l=JSON.parse(new TextDecoder().decode(d[a]));this.addCustomWord(l),H.saveWord(l,this._refsStorageKey)}catch{console.warn(`[Mellon] failed to parse ref file: ${i[a].path}`)}for(const a of s)this._refEmbeddings.set(a.word_name,a.embeddings);console.info("[Mellon] init complete, loaded refs:",[...this._refEmbeddings.keys()])}async init(o){this._initPromise||(this._initPromise=this._init(o)),await this._initPromise}addCustomWord(o){if(!(Array.isArray(o.embeddings)&&o.embeddings.length>0))throw new Error("invalid ref file for : "+o.word_name);this._refEmbeddings.set(o.word_name,o.embeddings)}async start(){if(this._started)return;await this.init();let o;try{o=await navigator.mediaDevices.getUserMedia({audio:{noiseSuppression:!1,echoCancellation:!1,autoGainControl:!1,channelCount:1}})}catch{o=await navigator.mediaDevices.getUserMedia({audio:!0})}this._stream=o;const t=new AudioContext({sampleRate:16e3});this._audioCtx=t,await t.audioWorklet.addModule(this._audioProcessorPath);const s=t.createMediaStreamSource(o),n=new AudioWorkletNode(t,"audio-processor");n.port.onmessage=e=>{this._handleBuffer(e.data)},s.connect(n),n.connect(t.destination),this._started=!0}async stop(){if(this._started=!1,this._audioCtx&&(await this._audioCtx.close(),this._audioCtx=null),this._stream){for(const o of this._stream.getTracks())o.stop();this._stream=null}}async _handleBuffer(o){if(this._inferring)return;const t=Date.now();if(!(t-this._lastInferenceAt<300)){this._lastInferenceAt=t,this._inferring=!0;try{const[s,n]=await Promise.all([et(this._wasmPaths,this._ortCdnUrl),mt(this._modelPath,this._wasmPaths,this._ortCdnUrl)]),e=this._audioUtils.logfbank(o),i=new s.Tensor("float32",e,[1,1,149,64]),r=await n.run({input:i}),h=r[Object.keys(r)[0]].data;let c=!1;for(const d of this._commands){if(c)break;for(const a of d.triggers){const l=this._refEmbeddings.get(a.name);if(!l)continue;const _=this._audioUtils.maxCosineSim(h,l);if(_>=this._threshold&&t-this._lastMatchAt>2e3){this._lastMatchAt=t,console.info(`[Mellon] match: "${a}" sim=${_.toFixed(3)}`),typeof d.onMatch=="function"&&d.onMatch(a.name,_),c=!0;break}}}}catch(s){console.error("[Mellon] inference error:",s)}finally{this._inferring=!1}}}}class Wt{constructor(o,t){this._config={},this._samples=[],this._wordName=o,this._config.modelPath=(t==null?void 0:t.modelPath)||rt,this._config.wasmPaths=(t==null?void 0:t.wasmPaths)||Z,this._config.ortCdnUrl=(t==null?void 0:t.ortCdnUrl)||q,this._audioUtils=(t==null?void 0:t.audioUtils)??new ft}async recordSample(){const o=await navigator.mediaDevices.getUserMedia({audio:!0}),t=new AudioContext({sampleRate:16e3}),s=await new Promise((i,r)=>{const h=new MediaRecorder(o),c=[];h.ondataavailable=d=>{d.data.size>0&&c.push(d.data)},h.onstop=async()=>{var d;for(const a of o.getTracks())a.stop();try{const l=await new Blob(c,{type:((d=c[0])==null?void 0:d.type)||"audio/webm"}).arrayBuffer(),_=await t.decodeAudioData(l);await t.close(),i(_.getChannelData(0).slice())}catch(a){r(a)}},h.start(),setTimeout(()=>{try{h.stop()}catch{}},1500)}),n=24e3,e=new Float32Array(n);return e.set(s.slice(0,n)),this._samples.push(e),this._samples.length}deleteSample(o){if(o<0||o>=this._samples.length)throw new RangeError(`index ${o} out of bounds (${this._samples.length} samples)`);return this._samples.splice(o,1),this._samples.length}async generateRef(){const[o,t]=await Promise.all([et(this._config.wasmPaths,this._config.ortCdnUrl),mt(this._config.modelPath,this._config.wasmPaths,this._config.ortCdnUrl)]),s=[];for(const n of this._samples){const e=this._audioUtils.logfbank(n),i=new o.Tensor("float32",e,[1,1,149,64]),r=await t.run({input:i}),h=Array.from(r[Object.keys(r)[0]].data);s.push(h)}return{word_name:this._wordName,model_type:"resnet_50_arc",embeddings:s}}}const Z="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/",q="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs",rt="https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx",wt="https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js",X="mellon-refs",gt="mellon-threshold";exports.AudioUtils=ft;exports.DEFAULT_AUDIO_PROCESSOR_PATH=wt;exports.DEFAULT_MODEL_PATH=rt;exports.DEFAULT_ORT_CDN_URL=q;exports.DEFAULT_REFS_STORAGE_KEY=X;exports.DEFAULT_THRESHOLD_STORAGE_KEY=gt;exports.DEFAULT_WASM_PATHS=Z;exports.Detector=Lt;exports.EnrollmentSession=Wt;exports.Storage=H;

package/dist/mellon.mjs

@@ -1,32 +1,32 @@
 function Rt(m) {
   return m && m.__esModule && Object.prototype.hasOwnProperty.call(m, "default") ? m.default : m;
 }
-var lt, vt;
+var dt, vt;
 function Ut() {
-  if (vt) return lt;
+  if (vt) return dt;
   vt = 1;
   function m(o) {
     if (this.size = o | 0, this.size <= 1 || (this.size & this.size - 1) !== 0)
       throw new Error("FFT size must be a power of two and bigger than 1");
     this._csize = o << 1;
     for (var t = new Array(this.size * 2), s = 0; s < t.length; s += 2) {
-      const h = Math.PI * s / this.size;
-      t[s] = Math.cos(h), t[s + 1] = -Math.sin(h);
+      const c = Math.PI * s / this.size;
+      t[s] = Math.cos(c), t[s + 1] = -Math.sin(c);
     }
     this.table = t;
     for (var n = 0, e = 1; this.size > e; e <<= 1)
       n++;
     this._width = n % 2 === 0 ? n - 1 : n, this._bitrev = new Array(1 << this._width);
-    for (var c = 0; c < this._bitrev.length; c++) {
-      this._bitrev[c] = 0;
+    for (var i = 0; i < this._bitrev.length; i++) {
+      this._bitrev[i] = 0;
       for (var r = 0; r < this._width; r += 2) {
-        var l = this._width - r - 2;
-        this._bitrev[c] |= (c >>> r & 3) << l;
+        var h = this._width - r - 2;
+        this._bitrev[i] |= (i >>> r & 3) << h;
       }
     }
     this._out = null, this._data = null, this._inv = 0;
   }
-  return lt = m, m.prototype.fromComplexArray = function(t, s) {
+  return dt = m, m.prototype.fromComplexArray = function(t, s) {
     for (var n = s || new Array(t.length >>> 1), e = 0; e < t.length; e += 2)
       n[e >>> 1] = t[e];
     return n;
@@ -58,70 +58,70 @@ function Ut() {
       t[n] /= this.size;
     this._out = null, this._data = null;
   }, m.prototype._transform4 = function() {
-    var t = this._out, s = this._csize, n = this._width, e = 1 << n, c = s / e << 1, r, l, h = this._bitrev;
-    if (c === 4)
-      for (r = 0, l = 0; r < s; r += c, l++) {
-        const u = h[l];
+    var t = this._out, s = this._csize, n = this._width, e = 1 << n, i = s / e << 1, r, h, c = this._bitrev;
+    if (i === 4)
+      for (r = 0, h = 0; r < s; r += i, h++) {
+        const u = c[h];
         this._singleTransform2(r, u, e);
       }
     else
-      for (r = 0, l = 0; r < s; r += c, l++) {
-        const u = h[l];
+      for (r = 0, h = 0; r < s; r += i, h++) {
+        const u = c[h];
         this._singleTransform4(r, u, e);
       }
-    var i = this._inv ? -1 : 1, a = this.table;
+    var d = this._inv ? -1 : 1, a = this.table;
     for (e >>= 2; e >= 2; e >>= 2) {
-      c = s / e << 1;
-      var d = c >>> 2;
-      for (r = 0; r < s; r += c)
-        for (var _ = r + d, v = r, f = 0; v < _; v += 2, f += e) {
-          const u = v, p = u + d, w = p + d, g = w + d, b = t[u], T = t[u + 1], F = t[p], y = t[p + 1], A = t[w], C = t[w + 1], M = t[g], S = t[g + 1], P = b, R = T, U = a[f], E = i * a[f + 1], D = F * U - y * E, x = F * E + y * U, B = a[2 * f], N = i * a[2 * f + 1], j = A * B - C * N, H = A * N + C * B, K = a[3 * f], $ = i * a[3 * f + 1], L = M * K - S * $, J = M * $ + S * K, G = P + j, z = R + H, I = P - j, Y = R - H, O = D + L, W = x + J, k = i * (D - L), Q = i * (x - J), X = G + O, et = z + W, rt = G - O, ot = z - W, nt = I + Q, at = Y - k, it = I - Q, ct = Y + k;
-          t[u] = X, t[u + 1] = et, t[p] = nt, t[p + 1] = at, t[w] = rt, t[w + 1] = ot, t[g] = it, t[g + 1] = ct;
+      i = s / e << 1;
+      var l = i >>> 2;
+      for (r = 0; r < s; r += i)
+        for (var _ = r + l, v = r, f = 0; v < _; v += 2, f += e) {
+          const u = v, p = u + l, w = p + l, g = w + l, b = t[u], T = t[u + 1], F = t[p], y = t[p + 1], A = t[w], C = t[w + 1], M = t[g], S = t[g + 1], P = b, R = T, U = a[f], E = d * a[f + 1], D = F * U - y * E, x = F * E + y * U, B = a[2 * f], N = d * a[2 * f + 1], j = A * B - C * N, H = A * N + C * B, K = a[3 * f], $ = d * a[3 * f + 1], L = M * K - S * $, J = M * $ + S * K, G = P + j, z = R + H, I = P - j, Y = R - H, O = D + L, W = x + J, k = d * (D - L), Q = d * (x - J), X = G + O, rt = z + W, ot = G - O, nt = z - W, at = I + Q, it = Y - k, ct = I - Q, lt = Y + k;
+          t[u] = X, t[u + 1] = rt, t[p] = at, t[p + 1] = it, t[w] = ot, t[w + 1] = nt, t[g] = ct, t[g + 1] = lt;
         }
     }
   }, m.prototype._singleTransform2 = function(t, s, n) {
-    const e = this._out, c = this._data, r = c[s], l = c[s + 1], h = c[s + n], i = c[s + n + 1], a = r + h, d = l + i, _ = r - h, v = l - i;
-    e[t] = a, e[t + 1] = d, e[t + 2] = _, e[t + 3] = v;
+    const e = this._out, i = this._data, r = i[s], h = i[s + 1], c = i[s + n], d = i[s + n + 1], a = r + c, l = h + d, _ = r - c, v = h - d;
+    e[t] = a, e[t + 1] = l, e[t + 2] = _, e[t + 3] = v;
   }, m.prototype._singleTransform4 = function(t, s, n) {
-    const e = this._out, c = this._data, r = this._inv ? -1 : 1, l = n * 2, h = n * 3, i = c[s], a = c[s + 1], d = c[s + n], _ = c[s + n + 1], v = c[s + l], f = c[s + l + 1], u = c[s + h], p = c[s + h + 1], w = i + v, g = a + f, b = i - v, T = a - f, F = d + u, y = _ + p, A = r * (d - u), C = r * (_ - p), M = w + F, S = g + y, P = b + C, R = T - A, U = w - F, E = g - y, D = b - C, x = T + A;
+    const e = this._out, i = this._data, r = this._inv ? -1 : 1, h = n * 2, c = n * 3, d = i[s], a = i[s + 1], l = i[s + n], _ = i[s + n + 1], v = i[s + h], f = i[s + h + 1], u = i[s + c], p = i[s + c + 1], w = d + v, g = a + f, b = d - v, T = a - f, F = l + u, y = _ + p, A = r * (l - u), C = r * (_ - p), M = w + F, S = g + y, P = b + C, R = T - A, U = w - F, E = g - y, D = b - C, x = T + A;
     e[t] = M, e[t + 1] = S, e[t + 2] = P, e[t + 3] = R, e[t + 4] = U, e[t + 5] = E, e[t + 6] = D, e[t + 7] = x;
   }, m.prototype._realTransform4 = function() {
-    var t = this._out, s = this._csize, n = this._width, e = 1 << n, c = s / e << 1, r, l, h = this._bitrev;
-    if (c === 4)
-      for (r = 0, l = 0; r < s; r += c, l++) {
-        const ht = h[l];
+    var t = this._out, s = this._csize, n = this._width, e = 1 << n, i = s / e << 1, r, h, c = this._bitrev;
+    if (i === 4)
+      for (r = 0, h = 0; r < s; r += i, h++) {
+        const ht = c[h];
         this._singleRealTransform2(r, ht >>> 1, e >>> 1);
       }
     else
-      for (r = 0, l = 0; r < s; r += c, l++) {
-        const ht = h[l];
+      for (r = 0, h = 0; r < s; r += i, h++) {
+        const ht = c[h];
         this._singleRealTransform4(r, ht >>> 1, e >>> 1);
       }
-    var i = this._inv ? -1 : 1, a = this.table;
+    var d = this._inv ? -1 : 1, a = this.table;
     for (e >>= 2; e >= 2; e >>= 2) {
-      c = s / e << 1;
-      var d = c >>> 1, _ = d >>> 1, v = _ >>> 1;
-      for (r = 0; r < s; r += c)
+      i = s / e << 1;
+      var l = i >>> 1, _ = l >>> 1, v = _ >>> 1;
+      for (r = 0; r < s; r += i)
         for (var f = 0, u = 0; f <= v; f += 2, u += e) {
-          var p = r + f, w = p + _, g = w + _, b = g + _, T = t[p], F = t[p + 1], y = t[w], A = t[w + 1], C = t[g], M = t[g + 1], S = t[b], P = t[b + 1], R = T, U = F, E = a[u], D = i * a[u + 1], x = y * E - A * D, B = y * D + A * E, N = a[2 * u], j = i * a[2 * u + 1], H = C * N - M * j, K = C * j + M * N, $ = a[3 * u], L = i * a[3 * u + 1], J = S * $ - P * L, G = S * L + P * $, z = R + H, I = U + K, Y = R - H, O = U - K, W = x + J, k = B + G, Q = i * (x - J), X = i * (B - G), et = z + W, rt = I + k, ot = Y + X, nt = O - Q;
-          if (t[p] = et, t[p + 1] = rt, t[w] = ot, t[w + 1] = nt, f === 0) {
-            var at = z - W, it = I - k;
-            t[g] = at, t[g + 1] = it;
+          var p = r + f, w = p + _, g = w + _, b = g + _, T = t[p], F = t[p + 1], y = t[w], A = t[w + 1], C = t[g], M = t[g + 1], S = t[b], P = t[b + 1], R = T, U = F, E = a[u], D = d * a[u + 1], x = y * E - A * D, B = y * D + A * E, N = a[2 * u], j = d * a[2 * u + 1], H = C * N - M * j, K = C * j + M * N, $ = a[3 * u], L = d * a[3 * u + 1], J = S * $ - P * L, G = S * L + P * $, z = R + H, I = U + K, Y = R - H, O = U - K, W = x + J, k = B + G, Q = d * (x - J), X = d * (B - G), rt = z + W, ot = I + k, nt = Y + X, at = O - Q;
+          if (t[p] = rt, t[p + 1] = ot, t[w] = nt, t[w + 1] = at, f === 0) {
+            var it = z - W, ct = I - k;
+            t[g] = it, t[g + 1] = ct;
             continue;
           }
           if (f !== v) {
-            var ct = Y, wt = -O, gt = z, yt = -I, bt = -i * X, Ft = -i * Q, Tt = -i * k, At = -i * W, Ct = ct + bt, Mt = wt + Ft, St = gt + At, Pt = yt - Tt, ft = r + _ - f, ut = r + d - f;
+            var lt = Y, wt = -O, gt = z, yt = -I, bt = -d * X, Ft = -d * Q, Tt = -d * k, At = -d * W, Ct = lt + bt, Mt = wt + Ft, St = gt + At, Pt = yt - Tt, ft = r + _ - f, ut = r + l - f;
             t[ft] = Ct, t[ft + 1] = Mt, t[ut] = St, t[ut + 1] = Pt;
           }
         }
     }
   }, m.prototype._singleRealTransform2 = function(t, s, n) {
-    const e = this._out, c = this._data, r = c[s], l = c[s + n], h = r + l, i = r - l;
-    e[t] = h, e[t + 1] = 0, e[t + 2] = i, e[t + 3] = 0;
+    const e = this._out, i = this._data, r = i[s], h = i[s + n], c = r + h, d = r - h;
+    e[t] = c, e[t + 1] = 0, e[t + 2] = d, e[t + 3] = 0;
   }, m.prototype._singleRealTransform4 = function(t, s, n) {
-    const e = this._out, c = this._data, r = this._inv ? -1 : 1, l = n * 2, h = n * 3, i = c[s], a = c[s + n], d = c[s + l], _ = c[s + h], v = i + d, f = i - d, u = a + _, p = r * (a - _), w = v + u, g = f, b = -p, T = v - u, F = f, y = p;
+    const e = this._out, i = this._data, r = this._inv ? -1 : 1, h = n * 2, c = n * 3, d = i[s], a = i[s + n], l = i[s + h], _ = i[s + c], v = d + l, f = d - l, u = a + _, p = r * (a - _), w = v + u, g = f, b = -p, T = v - u, F = f, y = p;
     e[t] = w, e[t + 1] = 0, e[t + 2] = g, e[t + 3] = b, e[t + 4] = T, e[t + 5] = 0, e[t + 6] = F, e[t + 7] = y;
-  }, lt;
+  }, dt;
 }
 var Et = Ut();
 const Dt = /* @__PURE__ */ Rt(Et);
@@ -137,40 +137,40 @@ class pt {
   }
   _createMelFilterbank() {
     const t = this._sampleRate / 2, s = this._hzToMel(0), n = this._hzToMel(t), e = new Float32Array(this._nfilt + 2);
-    for (let h = 0; h < this._nfilt + 2; h++)
-      e[h] = s + h * (n - s) / (this._nfilt + 1);
-    const r = e.map((h) => this._melToHz(h)).map((h) => Math.floor((this._nfft + 1) * h / this._sampleRate)), l = [];
-    for (let h = 0; h < this._nfilt; h++) {
-      const i = new Float32Array(Math.floor(this._nfft / 2) + 1);
-      for (let a = r[h]; a < r[h + 1]; a++)
-        i[a] = (a - r[h]) / (r[h + 1] - r[h]);
-      for (let a = r[h + 1]; a < r[h + 2]; a++)
-        i[a] = (r[h + 2] - a) / (r[h + 2] - r[h + 1]);
-      l.push(i);
+    for (let c = 0; c < this._nfilt + 2; c++)
+      e[c] = s + c * (n - s) / (this._nfilt + 1);
+    const r = e.map((c) => this._melToHz(c)).map((c) => Math.floor((this._nfft + 1) * c / this._sampleRate)), h = [];
+    for (let c = 0; c < this._nfilt; c++) {
+      const d = new Float32Array(Math.floor(this._nfft / 2) + 1);
+      for (let a = r[c]; a < r[c + 1]; a++)
+        d[a] = (a - r[c]) / (r[c + 1] - r[c]);
+      for (let a = r[c + 1]; a < r[c + 2]; a++)
+        d[a] = (r[c + 2] - a) / (r[c + 2] - r[c + 1]);
+      h.push(d);
     }
-    return l;
+    return h;
   }
   /** Returns a flat Float32Array of shape [numFrames × nfilt]. */
   logfbank(o) {
-    const t = Math.floor(0.025 * this._sampleRate), s = Math.floor(0.01 * this._sampleRate), n = 1 + Math.ceil((o.length - t) / s), e = new Float32Array(n * this._nfilt), c = new Float32Array(this._nfft), r = this._fft.createComplexArray();
-    for (let l = 0; l < n; l++) {
-      const h = l * s;
-      c.fill(0);
-      for (let d = 0; d < t && h + d < o.length; d++)
-        c[d] = o[h + d];
-      const i = this._fft.toComplexArray(c, null);
-      this._fft.transform(r, i);
+    const t = Math.floor(0.025 * this._sampleRate), s = Math.floor(0.01 * this._sampleRate), n = 1 + Math.ceil((o.length - t) / s), e = new Float32Array(n * this._nfilt), i = new Float32Array(this._nfft), r = this._fft.createComplexArray();
+    for (let h = 0; h < n; h++) {
+      const c = h * s;
+      i.fill(0);
+      for (let l = 0; l < t && c + l < o.length; l++)
+        i[l] = o[c + l];
+      const d = this._fft.toComplexArray(i, null);
+      this._fft.transform(r, d);
       const a = new Float32Array(Math.floor(this._nfft / 2) + 1);
-      for (let d = 0; d < a.length; d++) {
-        const _ = r[2 * d], v = r[2 * d + 1];
-        a[d] = 1 / this._nfft * (_ * _ + v * v), a[d] === 0 && (a[d] = 1e-30);
+      for (let l = 0; l < a.length; l++) {
+        const _ = r[2 * l], v = r[2 * l + 1];
+        a[l] = 1 / this._nfft * (_ * _ + v * v), a[l] === 0 && (a[l] = 1e-30);
       }
-      for (let d = 0; d < this._nfilt; d++) {
+      for (let l = 0; l < this._nfilt; l++) {
         let _ = 0;
-        const v = this._melFilters[d];
+        const v = this._melFilters[l];
         for (let f = 0; f < a.length; f++)
           _ += a[f] * v[f];
-        _ === 0 && (_ = 1e-30), e[l * this._nfilt + d] = Math.log(_);
+        _ === 0 && (_ = 1e-30), e[h * this._nfilt + l] = Math.log(_);
       }
     }
     return e;
@@ -180,22 +180,22 @@ class pt {
     for (const n of t) {
       let e = 0;
       for (let r = 0; r < n.length; r++) e += o[r] * n[r];
-      const c = (e + 1) / 2;
-      c > s && (s = c);
+      const i = (e + 1) / 2;
+      i > s && (s = i);
     }
     return s;
   }
 }
-async function q(m = tt, o = st) {
+async function tt(m = st, o = et) {
   const t = await import(
     /* @vite-ignore */
     o
   );
   return t.env.wasm.wasmPaths = m, t.env.wasm.numThreads = 1, t;
 }
-let dt = null;
-async function _t(m = mt, o = tt, t = st, s) {
-  return dt || (dt = q(o, t).then(
+let Z = null;
+async function _t(m = mt, o = st, t = et, s) {
+  return Z || (Z = tt(o, t).then(
     (n) => s ? n.InferenceSession.create(new Uint8Array(s), {
       executionProviders: ["wasm"],
       graphOptimizationLevel: "all"
@@ -203,10 +203,13 @@ async function _t(m = mt, o = tt, t = st, s) {
       executionProviders: ["wasm"],
       graphOptimizationLevel: "all"
     })
-  )), dt;
+  )), Z;
+}
+function xt() {
+  return Z !== null;
 }
 class V {
-  static loadWords(o = Z) {
+  static loadWords(o = q) {
     try {
       const t = localStorage.getItem(o);
       return t ? JSON.parse(t) : [];
@@ -214,11 +217,11 @@ class V {
       return [];
     }
   }
-  static saveWord(o, t = Z) {
+  static saveWord(o, t = q) {
     const s = V.loadWords(t).filter((n) => n.word_name !== o.word_name);
     localStorage.setItem(t, JSON.stringify([...s, o]));
   }
-  static deleteWord(o, t = Z) {
+  static deleteWord(o, t = q) {
     try {
       const s = V.loadWords(t).filter((n) => n.word_name !== o);
       localStorage.setItem(t, JSON.stringify(s));
@@ -226,22 +229,22 @@ class V {
     }
   }
 }
-class It {
+class Wt {
   constructor(o, t) {
     this._started = !1, this._inferring = !1, this._audioCtx = null, this._stream = null, this._refEmbeddings = /* @__PURE__ */ new Map(), this._lastMatchAt = 0, this._lastInferenceAt = 0, this._initPromise = null;
     const {
-      refsStorageKey: s = Z,
-      thresholdStorageKey: n = zt,
-      wasmPaths: e = tt,
-      modelPath: c = mt,
-      audioProcessorPath: r = xt,
-      ortCdnUrl: l = st,
-      audioUtils: h = new pt()
+      refsStorageKey: s = q,
+      thresholdStorageKey: n = It,
+      wasmPaths: e = st,
+      modelPath: i = mt,
+      audioProcessorPath: r = zt,
+      ortCdnUrl: h = et,
+      audioUtils: c = new pt()
     } = t || {};
-    this._audioUtils = h, this._commands = o, this._refsStorageKey = s, this._thresholdStorageKey = n, this._audioProcessorPath = r, this._wasmPaths = e, this._modelPath = c, this._ortCdnUrl = l;
+    this._audioUtils = c, this._commands = o, this._refsStorageKey = s, this._thresholdStorageKey = n, this._audioProcessorPath = r, this._wasmPaths = e, this._modelPath = i, this._ortCdnUrl = h;
     try {
-      const i = localStorage.getItem(this._thresholdStorageKey);
-      this._threshold = i !== null ? Math.max(0, Math.min(1, Number(i))) : 0.65;
+      const d = localStorage.getItem(this._thresholdStorageKey);
+      this._threshold = d !== null ? Math.max(0, Math.min(1, Number(d))) : 0.65;
     } catch {
       this._threshold = 0.65;
     }
@@ -271,40 +274,44 @@ class It {
       const a = await n.arrayBuffer();
       return s.downloaded += a.byteLength, e || (s.total += a.byteLength), t == null || t(s.downloaded, s.total), a;
     }
-    const c = n.body.getReader(), r = [];
-    let l = 0;
+    const i = n.body.getReader(), r = [];
+    let h = 0;
     for (; ; ) {
-      const { done: a, value: d } = await c.read();
+      const { done: a, value: l } = await i.read();
       if (a) break;
-      r.push(d), l += d.length, s.downloaded += d.length, t == null || t(s.downloaded, s.total);
+      r.push(l), h += l.length, s.downloaded += l.length, t == null || t(s.downloaded, s.total);
     }
-    e || (s.total += l);
-    const h = new Uint8Array(l);
-    let i = 0;
+    e || (s.total += h);
+    const c = new Uint8Array(h);
+    let d = 0;
     for (const a of r)
-      h.set(a, i), i += a.length;
-    return h.buffer;
+      c.set(a, d), d += a.length;
+    return c.buffer;
   }
   async _init(o) {
-    const t = { downloaded: 0, total: 0 }, s = /* @__PURE__ */ new Set(), n = [];
-    for (const i of this._commands)
-      for (const a of i.triggers)
-        !s.has(a.name) && a.defaultRefPath && (s.add(a.name), n.push({ name: a.name, path: a.defaultRefPath }));
-    const e = q(this._wasmPaths, this._ortCdnUrl), [c, ...r] = await Promise.all([
-      this._trackFetch(this._modelPath, o, t),
-      ...n.map(({ path: i }) => this._trackFetch(i, o, t))
+    const t = { downloaded: 0, total: 0 }, s = V.loadWords(this._refsStorageKey), n = new Set(s.map((a) => a.word_name)), e = /* @__PURE__ */ new Set(), i = [];
+    for (const a of this._commands)
+      for (const l of a.triggers)
+        !e.has(l.name) && l.defaultRefPath && !n.has(l.name) && (e.add(l.name), i.push({ name: l.name, path: l.defaultRefPath }));
+    const r = tt(this._wasmPaths, this._ortCdnUrl), h = xt(), [c, ...d] = await Promise.all([
+      h ? Promise.resolve(null) : this._trackFetch(this._modelPath, o, t),
+      ...i.map(({ path: a }) => this._trackFetch(a, o, t))
     ]);
-    await e, await _t(this._modelPath, this._wasmPaths, this._ortCdnUrl, c);
-    const l = V.loadWords(this._refsStorageKey), h = new Set(l.map((i) => i.word_name));
-    for (let i = 0; i < n.length; i++)
+    await r, await _t(
+      this._modelPath,
+      this._wasmPaths,
+      this._ortCdnUrl,
+      h ? void 0 : c
+    );
+    for (let a = 0; a < i.length; a++)
       try {
-        const a = JSON.parse(new TextDecoder().decode(r[i]));
-        this.addCustomWord(a), h.has(a.word_name) || V.saveWord(a, this._refsStorageKey);
+        const l = JSON.parse(new TextDecoder().decode(d[a]));
+        this.addCustomWord(l), V.saveWord(l, this._refsStorageKey);
       } catch {
-        console.warn(`[Mellon] failed to parse ref file: ${n[i].path}`);
+        console.warn(`[Mellon] failed to parse ref file: ${i[a].path}`);
       }
-    for (const i of l)
-      this._refEmbeddings.set(i.word_name, i.embeddings);
+    for (const a of s)
+      this._refEmbeddings.set(a.word_name, a.embeddings);
     console.info("[Mellon] init complete, loaded refs:", [...this._refEmbeddings.keys()]);
   }
   /**
@@ -359,16 +366,16 @@ class It {
     if (!(t - this._lastInferenceAt < 300)) {
       this._lastInferenceAt = t, this._inferring = !0;
       try {
-        const [s, n] = await Promise.all([q(this._wasmPaths, this._ortCdnUrl), _t(this._modelPath, this._wasmPaths, this._ortCdnUrl)]), e = this._audioUtils.logfbank(o), c = new s.Tensor("float32", e, [1, 1, 149, 64]), r = await n.run({ input: c }), l = r[Object.keys(r)[0]].data;
-        let h = !1;
-        for (const i of this._commands) {
-          if (h) break;
-          for (const a of i.triggers) {
-            const d = this._refEmbeddings.get(a.name);
-            if (!d) continue;
-            const _ = this._audioUtils.maxCosineSim(l, d);
+        const [s, n] = await Promise.all([tt(this._wasmPaths, this._ortCdnUrl), _t(this._modelPath, this._wasmPaths, this._ortCdnUrl)]), e = this._audioUtils.logfbank(o), i = new s.Tensor("float32", e, [1, 1, 149, 64]), r = await n.run({ input: i }), h = r[Object.keys(r)[0]].data;
+        let c = !1;
+        for (const d of this._commands) {
+          if (c) break;
+          for (const a of d.triggers) {
+            const l = this._refEmbeddings.get(a.name);
+            if (!l) continue;
+            const _ = this._audioUtils.maxCosineSim(h, l);
             if (_ >= this._threshold && t - this._lastMatchAt > 2e3) {
-              this._lastMatchAt = t, console.info(`[Mellon] match: "${a}" sim=${_.toFixed(3)}`), typeof i.onMatch == "function" && i.onMatch(a.name, _), h = !0;
+              this._lastMatchAt = t, console.info(`[Mellon] match: "${a}" sim=${_.toFixed(3)}`), typeof d.onMatch == "function" && d.onMatch(a.name, _), c = !0;
               break;
             }
           }
@@ -381,28 +388,28 @@ class It {
     }
   }
 }
-class Wt {
+class kt {
   constructor(o, t) {
-    this._config = {}, this._samples = [], this._wordName = o, this._config.modelPath = (t == null ? void 0 : t.modelPath) || mt, this._config.wasmPaths = (t == null ? void 0 : t.wasmPaths) || tt, this._config.ortCdnUrl = (t == null ? void 0 : t.ortCdnUrl) || st, this._audioUtils = (t == null ? void 0 : t.audioUtils) ?? new pt();
+    this._config = {}, this._samples = [], this._wordName = o, this._config.modelPath = (t == null ? void 0 : t.modelPath) || mt, this._config.wasmPaths = (t == null ? void 0 : t.wasmPaths) || st, this._config.ortCdnUrl = (t == null ? void 0 : t.ortCdnUrl) || et, this._audioUtils = (t == null ? void 0 : t.audioUtils) ?? new pt();
   }
   /** Records 1.5 s of audio, stores the decoded PCM, returns new sample count. */
   async recordSample() {
-    const o = await navigator.mediaDevices.getUserMedia({ audio: !0 }), t = new AudioContext({ sampleRate: 16e3 }), s = await new Promise((c, r) => {
-      const l = new MediaRecorder(o), h = [];
-      l.ondataavailable = (i) => {
-        i.data.size > 0 && h.push(i.data);
-      }, l.onstop = async () => {
-        var i;
+    const o = await navigator.mediaDevices.getUserMedia({ audio: !0 }), t = new AudioContext({ sampleRate: 16e3 }), s = await new Promise((i, r) => {
+      const h = new MediaRecorder(o), c = [];
+      h.ondataavailable = (d) => {
+        d.data.size > 0 && c.push(d.data);
+      }, h.onstop = async () => {
+        var d;
         for (const a of o.getTracks()) a.stop();
         try {
-          const d = await new Blob(h, { type: ((i = h[0]) == null ? void 0 : i.type) || "audio/webm" }).arrayBuffer(), _ = await t.decodeAudioData(d);
-          await t.close(), c(_.getChannelData(0).slice());
+          const l = await new Blob(c, { type: ((d = c[0]) == null ? void 0 : d.type) || "audio/webm" }).arrayBuffer(), _ = await t.decodeAudioData(l);
+          await t.close(), i(_.getChannelData(0).slice());
         } catch (a) {
           r(a);
         }
-      }, l.start(), setTimeout(() => {
+      }, h.start(), setTimeout(() => {
         try {
-          l.stop();
+          h.stop();
         } catch {
         }
       }, 1500);
@@ -417,24 +424,24 @@ class Wt {
   }
   /** Runs ONNX inference on every recorded sample to produce reference embeddings. */
   async generateRef() {
-    const [o, t] = await Promise.all([q(this._config.wasmPaths, this._config.ortCdnUrl), _t(this._config.modelPath, this._config.wasmPaths, this._config.ortCdnUrl)]), s = [];
+    const [o, t] = await Promise.all([tt(this._config.wasmPaths, this._config.ortCdnUrl), _t(this._config.modelPath, this._config.wasmPaths, this._config.ortCdnUrl)]), s = [];
     for (const n of this._samples) {
-      const e = this._audioUtils.logfbank(n), c = new o.Tensor("float32", e, [1, 1, 149, 64]), r = await t.run({ input: c }), l = Array.from(r[Object.keys(r)[0]].data);
-      s.push(l);
+      const e = this._audioUtils.logfbank(n), i = new o.Tensor("float32", e, [1, 1, 149, 64]), r = await t.run({ input: i }), h = Array.from(r[Object.keys(r)[0]].data);
+      s.push(h);
     }
     return { word_name: this._wordName, model_type: "resnet_50_arc", embeddings: s };
   }
 }
-const tt = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/", st = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs", mt = "https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx", xt = "https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js", Z = "mellon-refs", zt = "mellon-threshold";
+const st = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/", et = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs", mt = "https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx", zt = "https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js", q = "mellon-refs", It = "mellon-threshold";
 export {
   pt as AudioUtils,
-  xt as DEFAULT_AUDIO_PROCESSOR_PATH,
+  zt as DEFAULT_AUDIO_PROCESSOR_PATH,
   mt as DEFAULT_MODEL_PATH,
-  st as DEFAULT_ORT_CDN_URL,
-  Z as DEFAULT_REFS_STORAGE_KEY,
-  zt as DEFAULT_THRESHOLD_STORAGE_KEY,
-  tt as DEFAULT_WASM_PATHS,
-  It as Detector,
-  Wt as EnrollmentSession,
+  et as DEFAULT_ORT_CDN_URL,
+  q as DEFAULT_REFS_STORAGE_KEY,
+  It as DEFAULT_THRESHOLD_STORAGE_KEY,
+  st as DEFAULT_WASM_PATHS,
+  Wt as Detector,
+  kt as EnrollmentSession,
   V as Storage
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mellon",
-  "version": "0.0.20",
+  "version": "0.0.22",
   "description": "Offline, in-browser voice commands powered by EfficientWord-Net (ResNet-50 ArcFace).",
   "type": "module",
   "main": "./dist/mellon.cjs",