@onjmin/koe 1.0.3 → 1.0.4

package/dist/koe-convert.js

@@ -2,133 +2,35 @@
 #!/usr/bin/env node
 
 // src/converter/cli.ts
-import { readdir, readFile, writeFile, mkdir } from "fs/promises";
-import { join, dirname, basename, resolve } from "path";
+import { mkdir, readdir, readFile, writeFile } from "fs/promises";
+import { basename, dirname, join, resolve, sep } from "path";
 
-// src/converter/parse-oto.ts
-function parseOto(content) {
-  const entries = [];
-  for (const raw of content.split(/\r?\n/)) {
-    const line = raw.trim();
-    if (!line || line.startsWith("#")) continue;
-    const eq = line.indexOf("=");
-    if (eq === -1) continue;
-    const wav = line.slice(0, eq).trim();
-    const parts = line.slice(eq + 1).split(",");
-    if (parts.length < 6) continue;
-    const [alias, offsetStr, consonantStr, cutoffStr, preStr, overlapStr] = parts;
-    const aliasStr = alias.trim() || wav.replace(/\.[^.]+$/, "");
-    const entry = {
-      wav,
-      alias: aliasStr,
-      offset: parseFloat(offsetStr) || 0,
-      consonant: parseFloat(consonantStr) || 0,
-      cutoff: parseFloat(cutoffStr) || 0,
-      pre: parseFloat(preStr) || 0,
-      overlap: parseFloat(overlapStr) || 0
-    };
-    if (!entry.alias) continue;
-    entries.push(entry);
-  }
-  return entries;
+// src/koe.ts
+var MAGIC = 1263486208;
+function packKoe(manifest, pcmParts) {
+  const json = new TextEncoder().encode(JSON.stringify(manifest));
+  const header = new ArrayBuffer(8);
+  const view = new DataView(header);
+  view.setUint32(0, MAGIC, false);
+  view.setUint32(4, json.byteLength, true);
+  return new Blob([header, json, ...pcmParts]);
 }
 
-// src/converter/wav.ts
-function parseWav(buf) {
-  const view = new DataView(buf);
-  const riff = readFourCC(view, 0);
-  if (riff !== "RIFF") throw new Error(`Not a RIFF file (got "${riff}")`);
-  let sampleRate = 0;
-  let channels = 0;
-  let bitsPerSample = 0;
-  let audioFormat = 1;
-  let dataOffset = 0;
-  let dataLength = 0;
-  let pos = 12;
-  while (pos < view.byteLength - 8) {
-    const id = readFourCC(view, pos);
-    const size = view.getUint32(pos + 4, true);
-    pos += 8;
-    if (id === "fmt ") {
-      audioFormat = view.getUint16(pos, true);
-      channels = view.getUint16(pos + 2, true);
-      sampleRate = view.getUint32(pos + 4, true);
-      bitsPerSample = view.getUint16(pos + 14, true);
-    } else if (id === "data") {
-      dataOffset = pos;
-      dataLength = size;
-      break;
-    }
-    pos += size + (size & 1);
-  }
-  if (!dataOffset) throw new Error("WAV has no data chunk");
-  if (!channels || !sampleRate) throw new Error("WAV fmt chunk missing");
-  const bytesPerSample = bitsPerSample >> 3;
-  const totalSamples = Math.floor(dataLength / bytesPerSample);
-  const samples = new Float32Array(totalSamples);
-  for (let i = 0; i < totalSamples; i++) {
-    const p = dataOffset + i * bytesPerSample;
-    if (audioFormat === 3) {
-      samples[i] = view.getFloat32(p, true);
-    } else if (bitsPerSample === 8) {
-      samples[i] = (view.getUint8(p) - 128) / 128;
-    } else if (bitsPerSample === 16) {
-      samples[i] = view.getInt16(p, true) / 32768;
-    } else if (bitsPerSample === 24) {
-      const lo = view.getUint8(p) | view.getUint8(p + 1) << 8;
-      let hi = view.getUint8(p + 2);
-      if (hi & 128) hi = hi | 4294967040;
-      samples[i] = (hi << 16 | lo) / 8388608;
-    }
-  }
-  return { sampleRate, channels, samples };
-}
-function toMono(wav) {
-  if (wav.channels === 1) return wav;
-  const len = wav.samples.length / wav.channels;
-  const out = new Float32Array(len);
-  for (let i = 0; i < len; i++) {
-    let sum = 0;
-    for (let c = 0; c < wav.channels; c++)
-      sum += wav.samples[i * wav.channels + c];
-    out[i] = sum / wav.channels;
-  }
-  return { sampleRate: wav.sampleRate, channels: 1, samples: out };
-}
-function resample(wav, targetRate) {
-  if (wav.sampleRate === targetRate) return wav;
-  const ratio = wav.sampleRate / targetRate;
-  const outLen = Math.floor(wav.samples.length / ratio);
-  const out = new Float32Array(outLen);
-  const src = wav.samples;
-  for (let i = 0; i < outLen; i++) {
-    const x = i * ratio;
-    const xi = Math.floor(x);
-    const frac = x - xi;
-    out[i] = (src[xi] ?? 0) + ((src[xi + 1] ?? 0) - (src[xi] ?? 0)) * frac;
-  }
-  return { sampleRate: targetRate, channels: 1, samples: out };
-}
-function toInt16(samples) {
-  const out = new Int16Array(samples.length);
-  for (let i = 0; i < samples.length; i++) {
-    out[i] = Math.round(Math.max(-1, Math.min(1, samples[i])) * 32767);
-  }
-  return out;
-}
-function normalizePcm(buf) {
-  const wav = parseWav(buf);
-  const mono = toMono(wav);
-  const resampled = resample(mono, 48e3);
-  return toInt16(resampled.samples);
+// src/converter/frq.ts
+function parseFrqAverageF0(buffer) {
+  if (buffer.byteLength < 20) 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;
 }
-function readFourCC(view, pos) {
-  return String.fromCharCode(
-    view.getUint8(pos),
-    view.getUint8(pos + 1),
-    view.getUint8(pos + 2),
-    view.getUint8(pos + 3)
-  );
+function frqFileName(wavName) {
+  const dot = wavName.lastIndexOf(".");
+  const base = dot >= 0 ? wavName.slice(0, dot) : wavName;
+  const ext = dot >= 0 ? wavName.slice(dot + 1) : "wav";
+  return `${base}_${ext}.frq`;
 }
 
 // src/converter/pitch.ts
@@ -258,32 +160,139 @@ function pack(inputs, referencePitch = 220) {
   return { manifest, bin };
 }
 
-// src/converter/frq.ts
-function parseFrqAverageF0(buffer) {
-  if (buffer.byteLength < 20) 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;
-}
-function frqFileName(wavName) {
-  const dot = wavName.lastIndexOf(".");
-  const base = dot >= 0 ? wavName.slice(0, dot) : wavName;
-  const ext = dot >= 0 ? wavName.slice(dot + 1) : "wav";
-  return `${base}_${ext}.frq`;
+// src/converter/parse-oto.ts
+function parseOto(content) {
+  const entries = [];
+  for (const raw of content.split(/\r?\n/)) {
+    const line = raw.trim();
+    if (!line || line.startsWith("#")) continue;
+    const eq = line.indexOf("=");
+    if (eq === -1) continue;
+    const wav = line.slice(0, eq).trim();
+    const parts = line.slice(eq + 1).split(",");
+    if (parts.length < 6) continue;
+    const [alias, offsetStr, consonantStr, cutoffStr, preStr, overlapStr] = parts;
+    const aliasStr = alias.trim() || wav.replace(/\.[^.]+$/, "");
+    const entry = {
+      wav,
+      alias: aliasStr,
+      offset: parseFloat(offsetStr) || 0,
+      consonant: parseFloat(consonantStr) || 0,
+      cutoff: parseFloat(cutoffStr) || 0,
+      pre: parseFloat(preStr) || 0,
+      overlap: parseFloat(overlapStr) || 0
+    };
+    if (!entry.alias) continue;
+    entries.push(entry);
+  }
+  return entries;
 }
 
-// src/koe.ts
-var MAGIC = 1263486208;
-function packKoe(manifest, pcmParts) {
-  const json = new TextEncoder().encode(JSON.stringify(manifest));
-  const header = new ArrayBuffer(8);
-  const view = new DataView(header);
-  view.setUint32(0, MAGIC, false);
-  view.setUint32(4, json.byteLength, true);
-  return new Blob([header, json, ...pcmParts]);
+// src/converter/wav.ts
+function parseWav(buf) {
+  const view = new DataView(buf);
+  const riff = readFourCC(view, 0);
+  if (riff !== "RIFF") throw new Error(`Not a RIFF file (got "${riff}")`);
+  let sampleRate = 0;
+  let channels = 0;
+  let bitsPerSample = 0;
+  let audioFormat = 1;
+  let dataOffset = 0;
+  let dataLength = 0;
+  let pos = 12;
+  while (pos < view.byteLength - 8) {
+    const id = readFourCC(view, pos);
+    const size = view.getUint32(pos + 4, true);
+    pos += 8;
+    if (id === "fmt ") {
+      audioFormat = view.getUint16(pos, true);
+      channels = view.getUint16(pos + 2, true);
+      sampleRate = view.getUint32(pos + 4, true);
+      bitsPerSample = view.getUint16(pos + 14, true);
+      if (audioFormat === 65534 && size >= 40) {
+        audioFormat = view.getUint16(pos + 24, true);
+      }
+    } else if (id === "data") {
+      dataOffset = pos;
+      dataLength = Math.min(size, view.byteLength - pos);
+      break;
+    }
+    pos += size + (size & 1);
+  }
+  if (!dataOffset) throw new Error("WAV has no data chunk");
+  if (!channels || !sampleRate) throw new Error("WAV fmt chunk missing");
+  const supported = audioFormat === 3 && bitsPerSample === 32 || audioFormat === 1 && (bitsPerSample === 8 || bitsPerSample === 16 || bitsPerSample === 24);
+  if (!supported) {
+    throw new Error(
+      `Unsupported WAV format ${audioFormat} / ${bitsPerSample}-bit (need PCM 8/16/24-bit or IEEE float 32-bit)`
+    );
+  }
+  const bytesPerSample = bitsPerSample >> 3;
+  const totalSamples = Math.floor(dataLength / bytesPerSample);
+  const samples = new Float32Array(totalSamples);
+  for (let i = 0; i < totalSamples; i++) {
+    const p = dataOffset + i * bytesPerSample;
+    if (audioFormat === 3) {
+      samples[i] = view.getFloat32(p, true);
+    } else if (bitsPerSample === 8) {
+      samples[i] = (view.getUint8(p) - 128) / 128;
+    } else if (bitsPerSample === 16) {
+      samples[i] = view.getInt16(p, true) / 32768;
+    } else if (bitsPerSample === 24) {
+      const lo = view.getUint8(p) | view.getUint8(p + 1) << 8;
+      let hi = view.getUint8(p + 2);
+      if (hi & 128) hi = hi | 4294967040;
+      samples[i] = (hi << 16 | lo) / 8388608;
+    }
+  }
+  return { sampleRate, channels, samples };
+}
+function toMono(wav) {
+  if (wav.channels === 1) return wav;
+  const len = wav.samples.length / wav.channels;
+  const out = new Float32Array(len);
+  for (let i = 0; i < len; i++) {
+    let sum = 0;
+    for (let c = 0; c < wav.channels; c++)
+      sum += wav.samples[i * wav.channels + c];
+    out[i] = sum / wav.channels;
+  }
+  return { sampleRate: wav.sampleRate, channels: 1, samples: out };
+}
+function resample(wav, targetRate) {
+  if (wav.sampleRate === targetRate) return wav;
+  const ratio = wav.sampleRate / targetRate;
+  const outLen = Math.floor(wav.samples.length / ratio);
+  const out = new Float32Array(outLen);
+  const src = wav.samples;
+  for (let i = 0; i < outLen; i++) {
+    const x = i * ratio;
+    const xi = Math.floor(x);
+    const frac = x - xi;
+    out[i] = (src[xi] ?? 0) + ((src[xi + 1] ?? 0) - (src[xi] ?? 0)) * frac;
+  }
+  return { sampleRate: targetRate, channels: 1, samples: out };
+}
+function toInt16(samples) {
+  const out = new Int16Array(samples.length);
+  for (let i = 0; i < samples.length; i++) {
+    out[i] = Math.round(Math.max(-1, Math.min(1, samples[i])) * 32767);
+  }
+  return out;
+}
+function normalizePcm(buf) {
+  const wav = parseWav(buf);
+  const mono = toMono(wav);
+  const resampled = resample(mono, 48e3);
+  return toInt16(resampled.samples);
+}
+function readFourCC(view, pos) {
+  return String.fromCharCode(
+    view.getUint8(pos),
+    view.getUint8(pos + 1),
+    view.getUint8(pos + 2),
+    view.getUint8(pos + 3)
+  );
 }
 
 // src/converter/cli.ts
@@ -292,6 +301,20 @@ if (!voiceDir) {
   process.stderr.write("Usage: koe-convert <voice-dir> [output-dir]\n");
   process.exit(1);
 }
+function resolveInside(dir, name) {
+  const root = resolve(dir);
+  const path = resolve(root, name);
+  if (path !== root && !path.startsWith(root + sep)) {
+    throw new Error(`path escapes voice bank directory: ${name}`);
+  }
+  return path;
+}
+function toArrayBuffer(buf) {
+  return buf.buffer.slice(
+    buf.byteOffset,
+    buf.byteOffset + buf.byteLength
+  );
+}
 async function findOtoFiles(dir) {
   const found = [];
   const entries = await readdir(dir, { withFileTypes: true, recursive: true });
@@ -314,17 +337,22 @@ async function main() {
   for (const otoPath of otoFiles) {
     const otoDir = dirname(otoPath);
     const rawBytes = await readFile(otoPath);
-    const content = new TextDecoder("shift_jis").decode(rawBytes);
+    const isUtf8Bom = rawBytes[0] === 239 && rawBytes[1] === 187 && rawBytes[2] === 191;
+    const content = new TextDecoder(isUtf8Bom ? "utf-8" : "shift_jis").decode(
+      rawBytes
+    );
     const entries = parseOto(content);
     for (const oto of entries) {
-      const wavPath = join(otoDir, oto.wav);
       try {
+        const wavPath = resolveInside(otoDir, oto.wav);
         const wavBytes = await readFile(wavPath);
-        const pcm = normalizePcm(wavBytes.buffer);
+        const pcm = normalizePcm(toArrayBuffer(wavBytes));
         let recordedPitch = pitchFromAliasSuffix(oto.alias) ?? 0;
         try {
-          const frqBytes = await readFile(join(otoDir, frqFileName(oto.wav)));
-          const avg = parseFrqAverageF0(frqBytes.buffer);
+          const frqBytes = await readFile(
+            resolveInside(otoDir, frqFileName(oto.wav))
+          );
+          const avg = parseFrqAverageF0(toArrayBuffer(frqBytes));
           if (avg) recordedPitch = avg;
         } catch {
         }
@@ -332,7 +360,7 @@ async function main() {
       } catch (err) {
         const msg = err instanceof Error ? err.message : String(err);
         process.stderr.write(
-          `[skip] ${oto.alias} (${basename(wavPath)}): ${msg}
+          `[skip] ${oto.alias} (${basename(oto.wav)}): ${msg}
 `
         );
         skipped++;
@@ -344,7 +372,7 @@ async function main() {
     process.exit(1);
   }
   const { manifest, bin } = pack(inputs);
-  const koeName = basename(resolve(voiceDir)) + ".koe";
+  const koeName = `${basename(resolve(voiceDir))}.koe`;
   const koe = packKoe(manifest, [bin]);
   const koeBytes = Buffer.from(await koe.arrayBuffer());
   await mkdir(outDir, { recursive: true });

package/dist/koe-worklet.js

@@ -1 +1 @@
-"use strict";(()=>{var m=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)):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:m,length:0,consonant:0,pre:0,overlap:0,readPos:0,stepRate:1,remaining:e.duration};let t=this.manifest.phonemes[e.phoneme],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 a=this.readSample(r.data,r.readPos);if(!this.next&&this.queue.length===0&&r.remaining<=480&&r.length>0&&(a*=r.remaining/480),this.next){let s=this.next,o=Math.max(s.pre/s.stepRate,480);if(r.remaining<=o){let h=Math.min(Math.max(s.overlap/s.stepRate,480),o),l=o-r.remaining,c=Math.min(1,l/h),p=this.readSample(s.data,s.readPos);a=a*(1-c)+p*c,this.advance(s)}}n[i]=a,this.advance(r),r.remaining--}return!0}};registerProcessor("koe-processor",u);})();
+"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);})();

package/package.json

@@ -2,7 +2,7 @@
   "author": "onjmin",
   "license": "MIT",
   "name": "@onjmin/koe",
-  "version": "1.0.3",
+  "version": "1.0.4",
   "description": "UTAU concatenative synthesis engine optimized for browser / mobile via WebAssembly + AudioWorklet",
   "type": "module",
   "main": "./dist/index.js",