@onjmin/koe 1.0.4 → 1.0.6

package/dist/koe-convert.js

@@ -17,14 +17,48 @@ function packKoe(manifest, pcmParts) {
 }
 
 // src/converter/frq.ts
-function parseFrqAverageF0(buffer) {
-  if (buffer.byteLength < 20) return null;
+var HEADER_SIZE = 40;
+function parseFrq(buffer) {
+  if (buffer.byteLength < HEADER_SIZE) return null;
   const view = new DataView(buffer);
   let header = "";
   for (let i = 0; i < 8; i++) header += String.fromCharCode(view.getUint8(i));
   if (header !== "FREQ0003") return null;
-  const avg = view.getFloat64(12, true);
-  return Number.isFinite(avg) && avg > 0 ? avg : null;
+  const hopSize = view.getInt32(8, true);
+  const averageF0 = view.getFloat64(12, true);
+  const declared = view.getInt32(36, true);
+  const available = Math.floor((buffer.byteLength - HEADER_SIZE) / 16);
+  const length = Math.max(0, Math.min(declared, available));
+  const f0 = new Float64Array(length);
+  const amp = new Float64Array(length);
+  for (let i = 0; i < length; i++) {
+    const p = HEADER_SIZE + i * 16;
+    f0[i] = view.getFloat64(p, true);
+    amp[i] = view.getFloat64(p + 8, true);
+  }
+  return {
+    hopSize: hopSize > 0 ? hopSize : 256,
+    averageF0: Number.isFinite(averageF0) && averageF0 > 0 ? averageF0 : 0,
+    f0,
+    amp
+  };
+}
+function frqAverageF0InRange(frq, startMs, endMs, sourceRate) {
+  if (!(sourceRate > 0) || frq.f0.length === 0) return 0;
+  const perFrameMs = frq.hopSize / sourceRate * 1e3;
+  if (!(perFrameMs > 0)) return 0;
+  const first = Math.max(0, Math.floor(startMs / perFrameMs));
+  const last = Math.min(frq.f0.length - 1, Math.ceil(endMs / perFrameMs));
+  let num = 0;
+  let den = 0;
+  for (let i = first; i <= last; i++) {
+    const hz = frq.f0[i];
+    if (!(hz > 0)) continue;
+    const w = frq.amp[i] > 0 ? frq.amp[i] : 1;
+    num += hz * w;
+    den += w;
+  }
+  return den > 0 ? num / den : 0;
 }
 function frqFileName(wavName) {
   const dot = wavName.lastIndexOf(".");
@@ -112,11 +146,25 @@ function msToSamples(ms) {
   return Math.round(ms / 1e3 * TARGET_RATE);
 }
 var clamp = (v, lo, hi) => Math.max(lo, Math.min(hi, v));
+function otoRegion(pcmLength, oto) {
+  const start = clamp(msToSamples(oto.offset), 0, pcmLength);
+  const end = oto.cutoff < 0 ? clamp(start + msToSamples(-oto.cutoff), start, pcmLength) : clamp(pcmLength - msToSamples(oto.cutoff), start, pcmLength);
+  return { start, end };
+}
+function centerSlice(slice) {
+  const out = new Int16Array(slice.length);
+  if (slice.length === 0) return out;
+  let sum = 0;
+  for (let i = 0; i < slice.length; i++) sum += slice[i];
+  const dc = Math.round(sum / slice.length);
+  for (let i = 0; i < slice.length; i++) {
+    out[i] = clamp(slice[i] - dc, -32768, 32767);
+  }
+  return out;
+}
 function trimToOto(pcm, oto, recordedPitch = 0) {
-  const full = pcm.length;
-  const start = clamp(msToSamples(oto.offset), 0, full);
-  const end = oto.cutoff < 0 ? clamp(start + msToSamples(-oto.cutoff), start, full) : clamp(full - msToSamples(oto.cutoff), start, full);
-  const slice = pcm.subarray(start, end);
+  const { start, end } = otoRegion(pcm.length, oto);
+  const slice = centerSlice(pcm.subarray(start, end));
   const length = slice.length;
   const pre = clamp(msToSamples(oto.pre), 0, length);
   const overlap = clamp(msToSamples(oto.overlap), 0, length);
@@ -280,11 +328,11 @@ function toInt16(samples) {
   }
   return out;
 }
-function normalizePcm(buf) {
+function readWavPcm48k(buf) {
   const wav = parseWav(buf);
   const mono = toMono(wav);
   const resampled = resample(mono, 48e3);
-  return toInt16(resampled.samples);
+  return { pcm: toInt16(resampled.samples), sourceRate: wav.sampleRate };
 }
 function readFourCC(view, pos) {
   return String.fromCharCode(
@@ -346,14 +394,24 @@ async function main() {
       try {
         const wavPath = resolveInside(otoDir, oto.wav);
         const wavBytes = await readFile(wavPath);
-        const pcm = normalizePcm(toArrayBuffer(wavBytes));
+        const { pcm, sourceRate } = readWavPcm48k(toArrayBuffer(wavBytes));
         let recordedPitch = pitchFromAliasSuffix(oto.alias) ?? 0;
         try {
           const frqBytes = await readFile(
             resolveInside(otoDir, frqFileName(oto.wav))
           );
-          const avg = parseFrqAverageF0(toArrayBuffer(frqBytes));
-          if (avg) recordedPitch = avg;
+          const frq = parseFrq(toArrayBuffer(frqBytes));
+          if (frq) {
+            const { start, end } = otoRegion(pcm.length, oto);
+            const toMs = (samples) => samples / 48e3 * 1e3;
+            const local = frqAverageF0InRange(
+              frq,
+              toMs(start) + oto.pre,
+              toMs(end),
+              sourceRate
+            ) || frqAverageF0InRange(frq, toMs(start), toMs(end), sourceRate) || frq.averageF0;
+            if (local > 0) recordedPitch = local;
+          }
         } catch {
         }
         inputs.push({ oto, pcm, recordedPitch });

package/dist/koe-worklet.js

@@ -1 +1 @@
-"use strict";(()=>{var f=new Int16Array(0),u=class extends AudioWorkletProcessor{manifest=null;phonemes=new Map;queue=[];current=null;next=null;constructor(e){super(e),this.port.onmessage=t=>this.onMessage(t.data)}onMessage(e){e.type==="init"&&e.manifest?this.manifest=e.manifest:e.type==="phoneme"&&e.name&&e.buffer?this.phonemes.set(e.name,new Int16Array(e.buffer,0,Math.floor(e.buffer.byteLength/2))):e.type==="play"&&e.notes?this.queue.push(...e.notes):e.type==="stop"&&(this.queue=[],this.current=null,this.next=null)}dequeue(){let e=this.queue.shift();if(!e||!this.manifest)return null;if(e.phoneme==="")return{data:f,length:0,consonant:0,pre:0,overlap:0,readPos:0,stepRate:1,remaining:e.duration};let t=Object.hasOwn(this.manifest.phonemes,e.phoneme)?this.manifest.phonemes[e.phoneme]:void 0,n=this.phonemes.get(e.phoneme);if(!t||!n)return null;let i=t.pitch||this.manifest.referencePitch;return{data:n,length:t.length,consonant:t.consonant,pre:t.pre,overlap:t.overlap,readPos:0,stepRate:e.pitch/i,remaining:e.duration}}readSample(e,t){let n=t|0,i=t-n;return((e[n]??0)+((e[n+1]??0)-(e[n]??0))*i)/32768}advance(e){if(!(e.length<=0)&&(e.readPos+=e.stepRate,e.readPos>=e.length)){let t=e.length-e.consonant;e.readPos=t>0?e.consonant+(e.readPos-e.consonant)%t:e.length-1}}process(e,t){let n=t[0]?.[0];if(!n)return!0;for(let i=0;i<n.length;i++){if(!this.current||this.current.remaining<=0){let s=this.next!==null;if(this.current=this.next??this.dequeue(),this.next=null,!this.current){n[i]=0;continue}s||(this.current.readPos=this.current.pre)}let r=this.current;!this.next&&this.queue.length>0&&(this.next=this.dequeue());let o=this.readSample(r.data,r.readPos);if(!this.next&&this.queue.length===0&&r.remaining<=480&&r.length>0&&(o*=r.remaining/480),this.next){let s=this.next,a=Math.max(s.pre/s.stepRate,480);if(r.remaining<=a){let c=Math.min(Math.max(s.overlap/s.stepRate,480),a),l=a-r.remaining,h=Math.min(1,l/c),p=this.readSample(s.data,s.readPos);o=o*(1-h)+p*h,this.advance(s)}}n[i]=o,this.advance(r),r.remaining--}return!0}};registerProcessor("koe-processor",u);})();
+"use strict";(()=>{var P=new Int16Array(0),M=class extends AudioWorkletProcessor{manifest=null;phonemes=new Map;loops=new Map;queue=[];current=null;next=null;leadIn=!1;scratch=new Float32Array(256);constructor(e){super(e),this.port.onmessage=t=>this.onMessage(t.data)}onMessage(e){e.type==="init"&&e.manifest?this.manifest=e.manifest:e.type==="phoneme"&&e.name&&e.buffer?(this.phonemes.set(e.name,new Int16Array(e.buffer,0,Math.floor(e.buffer.byteLength/2))),this.loops.delete(e.name)):e.type==="play"&&e.notes?(this.leadIn=e.leadIn===!0,this.queue.push(...e.notes)):e.type==="stop"&&(this.queue=[],this.current=null,this.next=null)}dequeue(){let e=this.queue.shift();if(!e||!this.manifest)return null;if(e.phoneme==="")return{data:P,length:0,consonant:0,pre:0,overlap:0,readPos:0,stepRate:1,remaining:e.duration,duration:e.duration,period:0,lead:0,xfade:0,loop:{start:0,fade:0},fadeIn:0,aligned:!0};let t=Object.hasOwn(this.manifest.phonemes,e.phoneme)?this.manifest.phonemes[e.phoneme]:void 0,a=this.phonemes.get(e.phoneme);if(!t||!a)return null;let r=t.pitch||this.manifest.referencePitch||e.pitch||1;return{data:a,length:t.length,consonant:t.consonant,pre:t.pre,overlap:t.overlap,readPos:0,stepRate:e.pitch/r,remaining:e.duration,duration:e.duration,period:e.pitch>0?sampleRate/e.pitch:0,lead:0,xfade:0,loop:this.loopPoints(e.phoneme,a,t.length,t.consonant,sampleRate/r),fadeIn:0,aligned:!1}}readSample(e,t){let a=t|0,r=t-a;return((e[a]??0)+((e[a+1]??0)-(e[a]??0))*r)/32768}read(e){let t=this.readSample(e.data,e.readPos),a=e.loop.fade;if(a<=0)return t;let r=e.length-a;if(e.readPos<r)return t;let n=(e.readPos-r)/a,i=this.readSample(e.data,e.loop.start+(e.readPos-r));return t*(1-n)+i*n}wrap(e,t){let a=e.loop.fade,r=a>0?e.length-a-e.loop.start:e.length-e.consonant;return r<=0?e.length-1:(a>0?e.loop.start+a:e.consonant)+(t-e.length)%r}advance(e){e.length<=0||(e.readPos+=e.stepRate,e.readPos>=e.length&&(e.readPos=this.wrap(e,e.readPos)))}bestLag(e,t,a,r,n,i,s,o,h,d){let p=this.scratch,c=0;for(let l=0;l<s;l++){let m=this.readSample(e,t+l*a);p[l]=m,c+=m*m}if(c<=0)return 0;let f=0,u=-1/0;for(let l=o;l<=h;l+=d){let m=0,A=0;for(let b=0;b<s;b++){let v=this.readSample(r,n+l+b*i);m+=p[b]*v,A+=v*v}if(A<=0)continue;let g=m/Math.sqrt(A);(g>u||g===u&&Math.abs(l)<Math.abs(f))&&(u=g,f=l)}return f}loopPoints(e,t,a,r,n){let i=this.loops.get(e);if(i)return i;let s={start:r,fade:0},o=a-r;if(o>8&&n>=4&&Number.isFinite(n)){let h=Math.min(480,Math.floor(o/4)),d=a-h,p=Math.min(Math.round(n),Math.max(0,o-h*2-1)),c=r;if(p>0){let f=Math.min(h,256),u=Math.max(1,Math.round(p/64)),l=this.bestLag(t,d,1,t,r,1,f,0,p,u);u>1&&(l=this.bestLag(t,d,1,t,r,1,f,Math.max(0,l-u),Math.min(p,l+u),1)),c=r+l}c+h<a&&(s={start:c,fade:h})}return this.loops.set(e,s),s}prepareTransition(e,t){let a=Math.max(0,e.remaining*.5),r=Math.min(480,a),n=Math.max(t.pre/t.stepRate,r),i=e.length>0?Math.max(t.overlap/t.stepRate,r):r,s=Math.min(i,n);if(n>a&&n>0){let o=a/n;n*=o,s*=o}t.lead=n,t.xfade=s,t.readPos=Math.max(0,t.pre-n*t.stepRate),t.aligned=!1}alignPhase(e,t){if(t.aligned=!0,e.length<=0||t.length<=0||t.period<4||t.xfade<32)return;let a=Math.min(Math.round(t.period),Math.round(t.xfade),256);if(a<8)return;let r=t.period/2*t.stepRate,n=Math.max(1,Math.round(2*r/64)),i=this.bestLag(e.data,e.readPos,e.stepRate,t.data,t.readPos,t.stepRate,a,-r,r,n);n>1&&(i=this.bestLag(e.data,e.readPos,e.stepRate,t.data,t.readPos,t.stepRate,a,i-n,i+n,1));let s=t.readPos+i;t.readPos=s>=0?s:s+t.period*t.stepRate,t.readPos<0&&(t.readPos=0)}process(e,t){let a=t[0]?.[0];if(!a)return!0;for(let r=0;r<a.length;r++){if(!this.current||this.current.remaining<=0){let s=this.next!==null;if(this.current=this.next??this.dequeue(),this.next=null,!this.current){a[r]=0;continue}if(!s){let o=this.current;if(this.leadIn&&o.length>0){let h=o.pre/o.stepRate;o.readPos=0,o.remaining+=h,o.duration+=h}else o.readPos=o.pre;o.fadeIn=Math.min(480,o.duration)}}let n=this.current;!this.next&&this.queue.length>0&&(this.next=this.dequeue(),this.next&&this.prepareTransition(n,this.next));let i=this.read(n);if(n.fadeIn>0){let s=n.duration-n.remaining;s<n.fadeIn?i*=s/n.fadeIn:n.fadeIn=0}if(!this.next&&this.queue.length===0&&n.remaining<=480&&n.length>0&&(i*=n.remaining/480),this.next){let s=this.next;if(n.remaining<=s.lead){s.aligned||this.alignPhase(n,s);let o=s.lead-n.remaining,h=s.xfade>0?Math.min(1,o/s.xfade):1,d=this.read(s);i=i*(1-h)+d*h,this.advance(s)}}a[r]=i,this.advance(n),n.remaining--}return!0}};registerProcessor("koe-processor",M);})();

package/package.json

@@ -2,7 +2,7 @@
   "author": "onjmin",
   "license": "MIT",
   "name": "@onjmin/koe",
-  "version": "1.0.4",
+  "version": "1.0.6",
   "description": "UTAU concatenative synthesis engine optimized for browser / mobile via WebAssembly + AudioWorklet",
   "type": "module",
   "main": "./dist/index.js",
@@ -41,6 +41,9 @@
     "node": "24.16.0",
     "pnpm": "11.7.0"
   },
+  "dependencies": {
+    "fflate": "^0.8.3"
+  },
   "scripts": {
     "build": "tsup",
     "dev": "tsx src/dev-server.ts",