@onjmin/koe 1.0.6 → 1.0.8

package/dist/index.js

@@ -31,16 +31,16 @@ function frqAverageF0InRange(frq, startMs, endMs, sourceRate) {
   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 num2 = 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;
+    num2 += hz * w;
     den += w;
   }
-  return den > 0 ? num / den : 0;
+  return den > 0 ? num2 / den : 0;
 }
 function parseFrqAverageF0(buffer) {
   const frq = parseFrq(buffer);
@@ -490,7 +490,9 @@ var fdt = new u8(32);
 for (i = 0; i < 32; ++i)
   fdt[i] = 5;
 var i;
+var flm = /* @__PURE__ */ hMap(flt, 9, 0);
 var flrm = /* @__PURE__ */ hMap(flt, 9, 1);
+var fdm = /* @__PURE__ */ hMap(fdt, 5, 0);
 var fdrm = /* @__PURE__ */ hMap(fdt, 5, 1);
 var max = function(a) {
   var m = a[0];
@@ -686,7 +688,353 @@ var inflt = function(dat, st, buf, dict) {
   } while (!final);
   return bt != buf.length && noBuf ? slc(buf, 0, bt) : buf.subarray(0, bt);
 };
+var wbits = function(d, p, v) {
+  v <<= p & 7;
+  var o = p / 8 | 0;
+  d[o] |= v;
+  d[o + 1] |= v >> 8;
+};
+var wbits16 = function(d, p, v) {
+  v <<= p & 7;
+  var o = p / 8 | 0;
+  d[o] |= v;
+  d[o + 1] |= v >> 8;
+  d[o + 2] |= v >> 16;
+};
+var hTree = function(d, mb) {
+  var t = [];
+  for (var i = 0; i < d.length; ++i) {
+    if (d[i])
+      t.push({ s: i, f: d[i] });
+  }
+  var s = t.length;
+  var t2 = t.slice();
+  if (!s)
+    return { t: et, l: 0 };
+  if (s == 1) {
+    var v = new u8(t[0].s + 1);
+    v[t[0].s] = 1;
+    return { t: v, l: 1 };
+  }
+  t.sort(function(a, b) {
+    return a.f - b.f;
+  });
+  t.push({ s: -1, f: 25001 });
+  var l = t[0], r = t[1], i0 = 0, i1 = 1, i2 = 2;
+  t[0] = { s: -1, f: l.f + r.f, l, r };
+  while (i1 != s - 1) {
+    l = t[t[i0].f < t[i2].f ? i0++ : i2++];
+    r = t[i0 != i1 && t[i0].f < t[i2].f ? i0++ : i2++];
+    t[i1++] = { s: -1, f: l.f + r.f, l, r };
+  }
+  var maxSym = t2[0].s;
+  for (var i = 1; i < s; ++i) {
+    if (t2[i].s > maxSym)
+      maxSym = t2[i].s;
+  }
+  var tr = new u16(maxSym + 1);
+  var mbt = ln(t[i1 - 1], tr, 0);
+  if (mbt > mb) {
+    var i = 0, dt = 0;
+    var lft = mbt - mb, cst = 1 << lft;
+    t2.sort(function(a, b) {
+      return tr[b.s] - tr[a.s] || a.f - b.f;
+    });
+    for (; i < s; ++i) {
+      var i2_1 = t2[i].s;
+      if (tr[i2_1] > mb) {
+        dt += cst - (1 << mbt - tr[i2_1]);
+        tr[i2_1] = mb;
+      } else
+        break;
+    }
+    dt >>= lft;
+    while (dt > 0) {
+      var i2_2 = t2[i].s;
+      if (tr[i2_2] < mb)
+        dt -= 1 << mb - tr[i2_2]++ - 1;
+      else
+        ++i;
+    }
+    for (; i >= 0 && dt; --i) {
+      var i2_3 = t2[i].s;
+      if (tr[i2_3] == mb) {
+        --tr[i2_3];
+        ++dt;
+      }
+    }
+    mbt = mb;
+  }
+  return { t: new u8(tr), l: mbt };
+};
+var ln = function(n, l, d) {
+  return n.s == -1 ? Math.max(ln(n.l, l, d + 1), ln(n.r, l, d + 1)) : l[n.s] = d;
+};
+var lc = function(c) {
+  var s = c.length;
+  while (s && !c[--s])
+    ;
+  var cl = new u16(++s);
+  var cli = 0, cln = c[0], cls = 1;
+  var w = function(v) {
+    cl[cli++] = v;
+  };
+  for (var i = 1; i <= s; ++i) {
+    if (c[i] == cln && i != s)
+      ++cls;
+    else {
+      if (!cln && cls > 2) {
+        for (; cls > 138; cls -= 138)
+          w(32754);
+        if (cls > 2) {
+          w(cls > 10 ? cls - 11 << 5 | 28690 : cls - 3 << 5 | 12305);
+          cls = 0;
+        }
+      } else if (cls > 3) {
+        w(cln), --cls;
+        for (; cls > 6; cls -= 6)
+          w(8304);
+        if (cls > 2)
+          w(cls - 3 << 5 | 8208), cls = 0;
+      }
+      while (cls--)
+        w(cln);
+      cls = 1;
+      cln = c[i];
+    }
+  }
+  return { c: cl.subarray(0, cli), n: s };
+};
+var clen = function(cf, cl) {
+  var l = 0;
+  for (var i = 0; i < cl.length; ++i)
+    l += cf[i] * cl[i];
+  return l;
+};
+var wfblk = function(out, pos, dat) {
+  var s = dat.length;
+  var o = shft(pos + 2);
+  out[o] = s & 255;
+  out[o + 1] = s >> 8;
+  out[o + 2] = out[o] ^ 255;
+  out[o + 3] = out[o + 1] ^ 255;
+  for (var i = 0; i < s; ++i)
+    out[o + i + 4] = dat[i];
+  return (o + 4 + s) * 8;
+};
+var wblk = function(dat, out, final, syms, lf, df, eb, li, bs, bl, p) {
+  wbits(out, p++, final);
+  ++lf[256];
+  var _a2 = hTree(lf, 15), dlt = _a2.t, mlb = _a2.l;
+  var _b2 = hTree(df, 15), ddt = _b2.t, mdb = _b2.l;
+  var _c = lc(dlt), lclt = _c.c, nlc = _c.n;
+  var _d = lc(ddt), lcdt = _d.c, ndc = _d.n;
+  var lcfreq = new u16(19);
+  for (var i = 0; i < lclt.length; ++i)
+    ++lcfreq[lclt[i] & 31];
+  for (var i = 0; i < lcdt.length; ++i)
+    ++lcfreq[lcdt[i] & 31];
+  var _e = hTree(lcfreq, 7), lct = _e.t, mlcb = _e.l;
+  var nlcc = 19;
+  for (; nlcc > 4 && !lct[clim[nlcc - 1]]; --nlcc)
+    ;
+  var flen = bl + 5 << 3;
+  var ftlen = clen(lf, flt) + clen(df, fdt) + eb;
+  var dtlen = clen(lf, dlt) + clen(df, ddt) + eb + 14 + 3 * nlcc + clen(lcfreq, lct) + 2 * lcfreq[16] + 3 * lcfreq[17] + 7 * lcfreq[18];
+  if (bs >= 0 && flen <= ftlen && flen <= dtlen)
+    return wfblk(out, p, dat.subarray(bs, bs + bl));
+  var lm, ll, dm, dl;
+  wbits(out, p, 1 + (dtlen < ftlen)), p += 2;
+  if (dtlen < ftlen) {
+    lm = hMap(dlt, mlb, 0), ll = dlt, dm = hMap(ddt, mdb, 0), dl = ddt;
+    var llm = hMap(lct, mlcb, 0);
+    wbits(out, p, nlc - 257);
+    wbits(out, p + 5, ndc - 1);
+    wbits(out, p + 10, nlcc - 4);
+    p += 14;
+    for (var i = 0; i < nlcc; ++i)
+      wbits(out, p + 3 * i, lct[clim[i]]);
+    p += 3 * nlcc;
+    var lcts = [lclt, lcdt];
+    for (var it = 0; it < 2; ++it) {
+      var clct = lcts[it];
+      for (var i = 0; i < clct.length; ++i) {
+        var len = clct[i] & 31;
+        wbits(out, p, llm[len]), p += lct[len];
+        if (len > 15)
+          wbits(out, p, clct[i] >> 5 & 127), p += clct[i] >> 12;
+      }
+    }
+  } else {
+    lm = flm, ll = flt, dm = fdm, dl = fdt;
+  }
+  for (var i = 0; i < li; ++i) {
+    var sym = syms[i];
+    if (sym > 255) {
+      var len = sym >> 18 & 31;
+      wbits16(out, p, lm[len + 257]), p += ll[len + 257];
+      if (len > 7)
+        wbits(out, p, sym >> 23 & 31), p += fleb[len];
+      var dst = sym & 31;
+      wbits16(out, p, dm[dst]), p += dl[dst];
+      if (dst > 3)
+        wbits16(out, p, sym >> 5 & 8191), p += fdeb[dst];
+    } else {
+      wbits16(out, p, lm[sym]), p += ll[sym];
+    }
+  }
+  wbits16(out, p, lm[256]);
+  return p + ll[256];
+};
+var deo = /* @__PURE__ */ new i32([65540, 131080, 131088, 131104, 262176, 1048704, 1048832, 2114560, 2117632]);
 var et = /* @__PURE__ */ new u8(0);
+var dflt = function(dat, lvl, plvl, pre, post, st) {
+  var s = st.z || dat.length;
+  var o = new u8(pre + s + 5 * (1 + Math.ceil(s / 7e3)) + post);
+  var w = o.subarray(pre, o.length - post);
+  var lst = st.l;
+  var pos = (st.r || 0) & 7;
+  if (lvl) {
+    if (pos)
+      w[0] = st.r >> 3;
+    var opt = deo[lvl - 1];
+    var n = opt >> 13, c = opt & 8191;
+    var msk_1 = (1 << plvl) - 1;
+    var prev = st.p || new u16(32768), head = st.h || new u16(msk_1 + 1);
+    var bs1_1 = Math.ceil(plvl / 3), bs2_1 = 2 * bs1_1;
+    var hsh = function(i2) {
+      return (dat[i2] ^ dat[i2 + 1] << bs1_1 ^ dat[i2 + 2] << bs2_1) & msk_1;
+    };
+    var syms = new i32(25e3);
+    var lf = new u16(288), df = new u16(32);
+    var lc_1 = 0, eb = 0, i = st.i || 0, li = 0, wi = st.w || 0, bs = 0;
+    for (; i + 2 < s; ++i) {
+      var hv = hsh(i);
+      var imod = i & 32767, pimod = head[hv];
+      prev[imod] = pimod;
+      head[hv] = imod;
+      if (wi <= i) {
+        var rem = s - i;
+        if ((lc_1 > 7e3 || li > 24576) && (rem > 423 || !lst)) {
+          pos = wblk(dat, w, 0, syms, lf, df, eb, li, bs, i - bs, pos);
+          li = lc_1 = eb = 0, bs = i;
+          for (var j = 0; j < 286; ++j)
+            lf[j] = 0;
+          for (var j = 0; j < 30; ++j)
+            df[j] = 0;
+        }
+        var l = 2, d = 0, ch_1 = c, dif = imod - pimod & 32767;
+        if (rem > 2 && hv == hsh(i - dif)) {
+          var maxn = Math.min(n, rem) - 1;
+          var maxd = Math.min(32767, i);
+          var ml = Math.min(258, rem);
+          while (dif <= maxd && --ch_1 && imod != pimod) {
+            if (dat[i + l] == dat[i + l - dif]) {
+              var nl = 0;
+              for (; nl < ml && dat[i + nl] == dat[i + nl - dif]; ++nl)
+                ;
+              if (nl > l) {
+                l = nl, d = dif;
+                if (nl > maxn)
+                  break;
+                var mmd = Math.min(dif, nl - 2);
+                var md = 0;
+                for (var j = 0; j < mmd; ++j) {
+                  var ti = i - dif + j & 32767;
+                  var pti = prev[ti];
+                  var cd = ti - pti & 32767;
+                  if (cd > md)
+                    md = cd, pimod = ti;
+                }
+              }
+            }
+            imod = pimod, pimod = prev[imod];
+            dif += imod - pimod & 32767;
+          }
+        }
+        if (d) {
+          syms[li++] = 268435456 | revfl[l] << 18 | revfd[d];
+          var lin = revfl[l] & 31, din = revfd[d] & 31;
+          eb += fleb[lin] + fdeb[din];
+          ++lf[257 + lin];
+          ++df[din];
+          wi = i + l;
+          ++lc_1;
+        } else {
+          syms[li++] = dat[i];
+          ++lf[dat[i]];
+        }
+      }
+    }
+    for (i = Math.max(i, wi); i < s; ++i) {
+      syms[li++] = dat[i];
+      ++lf[dat[i]];
+    }
+    pos = wblk(dat, w, lst, syms, lf, df, eb, li, bs, i - bs, pos);
+    if (!lst) {
+      st.r = pos & 7 | w[pos / 8 | 0] << 3;
+      pos -= 7;
+      st.h = head, st.p = prev, st.i = i, st.w = wi;
+    }
+  } else {
+    for (var i = st.w || 0; i < s + lst; i += 65535) {
+      var e = i + 65535;
+      if (e >= s) {
+        w[pos / 8 | 0] = lst;
+        e = s;
+      }
+      pos = wfblk(w, pos + 1, dat.subarray(i, e));
+    }
+    st.i = s;
+  }
+  return slc(o, 0, pre + shft(pos) + post);
+};
+var crct = /* @__PURE__ */ (function() {
+  var t = new Int32Array(256);
+  for (var i = 0; i < 256; ++i) {
+    var c = i, k = 9;
+    while (--k)
+      c = (c & 1 && -306674912) ^ c >>> 1;
+    t[i] = c;
+  }
+  return t;
+})();
+var crc = function() {
+  var c = -1;
+  return {
+    p: function(d) {
+      var cr = c;
+      for (var i = 0; i < d.length; ++i)
+        cr = crct[cr & 255 ^ d[i]] ^ cr >>> 8;
+      c = cr;
+    },
+    d: function() {
+      return ~c;
+    }
+  };
+};
+var dopt = function(dat, opt, pre, post, st) {
+  if (!st) {
+    st = { l: 1 };
+    if (opt.dictionary) {
+      var dict = opt.dictionary.subarray(-32768);
+      var newDat = new u8(dict.length + dat.length);
+      newDat.set(dict);
+      newDat.set(dat, dict.length);
+      dat = newDat;
+      st.w = dict.length;
+    }
+  }
+  return dflt(dat, opt.level == null ? 6 : opt.level, opt.mem == null ? st.l ? Math.ceil(Math.max(8, Math.min(13, Math.log(dat.length))) * 1.5) : 20 : 12 + opt.mem, pre, post, st);
+};
+var mrg = function(a, b) {
+  var o = {};
+  for (var k in a)
+    o[k] = a[k];
+  for (var k in b)
+    o[k] = b[k];
+  return o;
+};
 var b2 = function(d, b) {
   return d[b] | d[b + 1] << 8;
 };
@@ -696,9 +1044,30 @@ var b4 = function(d, b) {
 var b8 = function(d, b) {
   return b4(d, b) + b4(d, b + 4) * 4294967296;
 };
+var wbytes = function(d, b, v) {
+  for (; v; ++b)
+    d[b] = v, v >>>= 8;
+};
+function deflateSync(data, opts) {
+  return dopt(data, opts || {}, 0, 0);
+}
 function inflateSync(data, opts) {
   return inflt(data, { i: 2 }, opts && opts.out, opts && opts.dictionary);
 }
+var fltn = function(d, p, t, o) {
+  for (var k in d) {
+    var val = d[k], n = p + k, op = o;
+    if (Array.isArray(val))
+      op = mrg(o, val[1]), val = val[0];
+    if (ArrayBuffer.isView(val))
+      t[n] = [val, op];
+    else {
+      t[n += "/"] = [new u8(0), op];
+      fltn(val, n, t, o);
+    }
+  }
+};
+var te = typeof TextEncoder != "undefined" && /* @__PURE__ */ new TextEncoder();
 var td = typeof TextDecoder != "undefined" && /* @__PURE__ */ new TextDecoder();
 var tds = 0;
 try {
@@ -722,6 +1091,39 @@ var dutf8 = function(d) {
       r += String.fromCharCode((c & 15) << 12 | (d[i++] & 63) << 6 | d[i++] & 63);
   }
 };
+function strToU8(str, latin1) {
+  if (latin1) {
+    var ar_1 = new u8(str.length);
+    for (var i = 0; i < str.length; ++i)
+      ar_1[i] = str.charCodeAt(i);
+    return ar_1;
+  }
+  if (te)
+    return te.encode(str);
+  var l = str.length;
+  var ar = new u8(str.length + (str.length >> 1));
+  var ai = 0;
+  var w = function(v) {
+    ar[ai++] = v;
+  };
+  for (var i = 0; i < l; ++i) {
+    if (ai + 5 > ar.length) {
+      var n = new u8(ai + 8 + (l - i << 1));
+      n.set(ar);
+      ar = n;
+    }
+    var c = str.charCodeAt(i);
+    if (c < 128 || latin1)
+      w(c);
+    else if (c < 2048)
+      w(192 | c >> 6), w(128 | c & 63);
+    else if (c > 55295 && c < 57344)
+      c = 65536 + (c & 1023 << 10) | str.charCodeAt(++i) & 1023, w(240 | c >> 18), w(128 | c >> 12 & 63), w(128 | c >> 6 & 63), w(128 | c & 63);
+    else
+      w(224 | c >> 12), w(128 | c >> 6 & 63), w(128 | c & 63);
+  }
+  return slc(ar, 0, ai);
+}
 function strFromU8(dat, latin1) {
   if (latin1) {
     var r = "";
@@ -764,6 +1166,107 @@ var z64hs = function(d, b, l, z, sc, su, off) {
   }
   return [sc, su, off, 0];
 };
+var exfl = function(ex) {
+  var le = 0;
+  if (ex) {
+    for (var k in ex) {
+      var l = ex[k].length;
+      if (l > 65535)
+        err(9);
+      le += l + 4;
+    }
+  }
+  return le;
+};
+var wzh = function(d, b, f, fn, u, c, ce, co) {
+  var fl2 = fn.length, ex = f.extra, col = co && co.length;
+  var exl = exfl(ex);
+  wbytes(d, b, ce != null ? 33639248 : 67324752), b += 4;
+  if (ce != null)
+    d[b++] = 20, d[b++] = f.os;
+  d[b] = 20, b += 2;
+  d[b++] = f.flag << 1 | (c < 0 && 8), d[b++] = u && 8;
+  d[b++] = f.compression & 255, d[b++] = f.compression >> 8;
+  var dt = new Date(f.mtime == null ? Date.now() : f.mtime), y = dt.getFullYear() - 1980;
+  if (y < 0 || y > 119)
+    err(10);
+  wbytes(d, b, y << 25 | dt.getMonth() + 1 << 21 | dt.getDate() << 16 | dt.getHours() << 11 | dt.getMinutes() << 5 | dt.getSeconds() >> 1), b += 4;
+  if (c != -1) {
+    wbytes(d, b, f.crc);
+    wbytes(d, b + 4, c < 0 ? -c - 2 : c);
+    wbytes(d, b + 8, f.size);
+  }
+  wbytes(d, b + 12, fl2);
+  wbytes(d, b + 14, exl), b += 16;
+  if (ce != null) {
+    wbytes(d, b, col);
+    wbytes(d, b + 6, f.attrs);
+    wbytes(d, b + 10, ce), b += 14;
+  }
+  d.set(fn, b);
+  b += fl2;
+  if (exl) {
+    for (var k in ex) {
+      var exf = ex[k], l = exf.length;
+      wbytes(d, b, +k);
+      wbytes(d, b + 2, l);
+      d.set(exf, b + 4), b += 4 + l;
+    }
+  }
+  if (col)
+    d.set(co, b), b += col;
+  return b;
+};
+var wzf = function(o, b, c, d, e) {
+  wbytes(o, b, 101010256);
+  wbytes(o, b + 8, c);
+  wbytes(o, b + 10, c);
+  wbytes(o, b + 12, d);
+  wbytes(o, b + 16, e);
+};
+function zipSync(data, opts) {
+  if (!opts)
+    opts = {};
+  var r = {};
+  var files = [];
+  fltn(data, "", r, opts);
+  var o = 0;
+  var tot = 0;
+  for (var fn in r) {
+    var _a2 = r[fn], file = _a2[0], p = _a2[1];
+    var compression = p.level == 0 ? 0 : 8;
+    var f = strToU8(fn), s = f.length;
+    var com = p.comment, m = com && strToU8(com), ms = m && m.length;
+    var exl = exfl(p.extra);
+    if (s > 65535)
+      err(11);
+    var d = compression ? deflateSync(file, p) : file, l = d.length;
+    var c = crc();
+    c.p(file);
+    files.push(mrg(p, {
+      size: file.length,
+      crc: c.d(),
+      c: d,
+      f,
+      m,
+      u: s != fn.length || m && com.length != ms,
+      o,
+      compression
+    }));
+    o += 30 + s + exl + l;
+    tot += 76 + 2 * (s + exl) + (ms || 0) + l;
+  }
+  var out = new u8(tot + 22), oe = o, cdl = tot - o;
+  for (var i = 0; i < files.length; ++i) {
+    var f = files[i];
+    wzh(out, f.o, f, f.f, f.u, f.c.length);
+    var badd = 30 + f.f.length + exfl(f.extra);
+    out.set(f.c, f.o + badd);
+    wzh(out, o, f, f.f, f.u, f.c.length, f.o, f.m), o += 16 + badd + (f.m ? f.m.length : 0);
+  }
+  wzf(out, o, files.length, cdl, oe);
+  return out;
+}
 function unzipSync(data, opts) {
   var files = {};
   var e = data.length - 22;
@@ -854,6 +1357,12 @@ function readUtf8Flags(buf) {
   }
   return flags;
 }
+function zipFiles(files) {
+  const packed = zipSync(files, { level: 6 });
+  const bytes = new Uint8Array(packed.length);
+  bytes.set(packed);
+  return new Blob([bytes], { type: "application/zip" });
+}
 
 // src/koe.ts
 var MAGIC = 1263486208;
@@ -1249,6 +1758,104 @@ var Worldline = class _Worldline {
    * @returns Float32 PCM, or null when `pcm` is shorter than
    *          {@link MIN_WORLDLINE_SAMPLES} (too short for stable F0 analysis).
    */
+  renderPhrase(params) {
+    const { units, pitch, gender = 0.5, tension = 0.5, breathiness = 0.5, voicing = 1, tempo = 120 } = params;
+    if (units.length === 0) return null;
+    const WL = this.wasm;
+    const FS = WORLDLINE_SAMPLE_RATE;
+    let totalMs = 0;
+    for (const u of units) {
+      const endMs = u.posMs + u.lengthMs;
+      if (endMs > totalMs) totalMs = endMs;
+    }
+    const ps = WL._PhraseSynthNew();
+    if (!ps) return null;
+    const pointersToFree = [];
+    for (const u of units) {
+      if (!u.pcm || u.pcm.length < MIN_WORLDLINE_SAMPLES) continue;
+      const reqPtr = WL._malloc(SYNTH_REQ_SIZE);
+      if (!reqPtr) continue;
+      pointersToFree.push(reqPtr);
+      const samplePtr = WL._malloc(u.pcm.length * 8);
+      if (!samplePtr) continue;
+      pointersToFree.push(samplePtr);
+      WL.HEAPF64.set(u.pcm, samplePtr >> 3);
+      const sv = (off, val, type) => WL.setValue(reqPtr + off, val, type);
+      sv(0, FS, "i32");
+      sv(4, u.pcm.length, "i32");
+      sv(8, samplePtr, "*");
+      sv(12, 0, "i32");
+      sv(16, 0, "*");
+      sv(20, 69, "i32");
+      sv(24, 100, "double");
+      sv(32, 0, "double");
+      sv(40, u.lengthMs, "double");
+      sv(48, u.consonantMs, "double");
+      const cutMs = u.cutMs ?? WL_FRAME_MS * 2;
+      sv(56, cutMs, "double");
+      sv(64, 100, "double");
+      sv(72, 0, "double");
+      sv(80, tempo, "double");
+      sv(88, 0, "i32");
+      sv(92, 0, "*");
+      sv(96, 0, "i32");
+      sv(100, 0, "i32");
+      sv(104, 100, "i32");
+      sv(108, 0, "i32");
+      sv(112, 0, "i32");
+      sv(116, 100, "i32");
+      WL._PhraseSynthAddRequest(ps, reqPtr, u.posMs, u.skipMs, u.lengthMs, u.fadeInMs, u.fadeOutMs, 0);
+    }
+    totalMs += WL_FRAME_MS * 2;
+    const nFrames = Math.ceil(totalMs / WL_FRAME_MS) + 4;
+    const f0Arr = new Float64Array(nFrames);
+    const gArr = new Float64Array(nFrames);
+    const tArr = new Float64Array(nFrames);
+    const bArr = new Float64Array(nFrames);
+    const vArr = new Float64Array(nFrames);
+    for (let i = 0; i < nFrames; i++) {
+      const tMs = i * WL_FRAME_MS;
+      f0Arr[i] = sampleCurve(tMs, pitch, totalMs);
+      gArr[i] = sampleCurve(tMs, gender, totalMs);
+      tArr[i] = sampleCurve(tMs, tension, totalMs);
+      bArr[i] = sampleCurve(tMs, breathiness, totalMs);
+      vArr[i] = sampleCurve(tMs, voicing, totalMs);
+    }
+    const f0Ptr = WL._malloc(nFrames * 8);
+    const gPtr = WL._malloc(nFrames * 8);
+    const tPtr = WL._malloc(nFrames * 8);
+    const bPtr = WL._malloc(nFrames * 8);
+    const vPtr = WL._malloc(nFrames * 8);
+    if (f0Ptr && gPtr && tPtr && bPtr && vPtr) {
+      WL.HEAPF64.set(f0Arr, f0Ptr >> 3);
+      WL.HEAPF64.set(gArr, gPtr >> 3);
+      WL.HEAPF64.set(tArr, tPtr >> 3);
+      WL.HEAPF64.set(bArr, bPtr >> 3);
+      WL.HEAPF64.set(vArr, vPtr >> 3);
+      WL._PhraseSynthSetCurves(ps, f0Ptr, gPtr, tPtr, bPtr, vPtr, nFrames, WL_FRAME_MS);
+    }
+    if (f0Ptr) WL._free(f0Ptr);
+    if (gPtr) WL._free(gPtr);
+    if (tPtr) WL._free(tPtr);
+    if (bPtr) WL._free(bPtr);
+    if (vPtr) WL._free(vPtr);
+    const yPtrPtr = WL._malloc(4);
+    let audio = null;
+    if (yPtrPtr) {
+      const outLen = WL._PhraseSynthSynth(ps, yPtrPtr, 0);
+      const yPtr = WL.getValue(yPtrPtr, "*");
+      if (outLen > 0 && yPtr) {
+        audio = new Float32Array(WL.HEAPF32.buffer, yPtr, outLen).slice();
+        WL._free(yPtr);
+      }
+      WL._free(yPtrPtr);
+    }
+    for (const ptr of pointersToFree) {
+      WL._free(ptr);
+    }
+    WL._PhraseSynthDelete(ps);
+    return audio;
+  }
   renderNote(params) {
     const {
       pcm,
@@ -1373,16 +1980,1302 @@ var Worldline = class _Worldline {
     return audio;
   }
 };
+
+// src/oto/frames.ts
+var RATE = 16e3;
+var HOP_MS = 2;
+var HOP = RATE * HOP_MS / 1e3;
+var FFT_SIZE = 512;
+var BINS = FFT_SIZE / 2;
+var PITCH_DECIM = 2;
+var PITCH_RATE = RATE / PITCH_DECIM;
+var PITCH_WIN = 512;
+var MIN_F0 = 70;
+var MAX_F0 = 600;
+var WINDOW_CENTRE_MS = FFT_SIZE / 2 / RATE * 1e3;
+function msToFrame(ms) {
+  return Math.round(ms / HOP_MS);
+}
+function framesToMs(frames) {
+  return frames * HOP_MS;
+}
+function frameTimeMs(frame) {
+  return frame * HOP_MS + WINDOW_CENTRE_MS;
+}
+function fft(re, im) {
+  const n = re.length;
+  for (let i = 1, j = 0; i < n; i++) {
+    let bit = n >> 1;
+    for (; j & bit; bit >>= 1) j ^= bit;
+    j ^= bit;
+    if (i < j) {
+      [re[i], re[j]] = [re[j], re[i]];
+      [im[i], im[j]] = [im[j], im[i]];
+    }
+  }
+  for (let len = 2; len <= n; len <<= 1) {
+    const ang = -2 * Math.PI / len;
+    const wRe = Math.cos(ang);
+    const wIm = Math.sin(ang);
+    for (let i = 0; i < n; i += len) {
+      let curRe = 1;
+      let curIm = 0;
+      for (let k = 0; k < len >> 1; k++) {
+        const aRe = re[i + k];
+        const aIm = im[i + k];
+        const bRe = re[i + k + (len >> 1)] * curRe - im[i + k + (len >> 1)] * curIm;
+        const bIm = re[i + k + (len >> 1)] * curIm + im[i + k + (len >> 1)] * curRe;
+        re[i + k] = aRe + bRe;
+        im[i + k] = aIm + bIm;
+        re[i + k + (len >> 1)] = aRe - bRe;
+        im[i + k + (len >> 1)] = aIm - bIm;
+        const nextRe = curRe * wRe - curIm * wIm;
+        curIm = curRe * wIm + curIm * wRe;
+        curRe = nextRe;
+      }
+    }
+  }
+}
+function quietestWindow(db) {
+  const averaged = smooth(db, Math.round(50 / HOP_MS));
+  let min = Infinity;
+  for (let i = 0; i < averaged.length; i++) {
+    if (averaged[i] < min) min = averaged[i];
+  }
+  return Number.isFinite(min) ? min : -120;
+}
+function percentile(values, p) {
+  const sorted = Float32Array.from(values).sort();
+  const i = Math.min(
+    sorted.length - 1,
+    Math.max(0, Math.round(p * (sorted.length - 1)))
+  );
+  return sorted[i];
+}
+function ncc(buf, at, win, lag) {
+  let r = 0;
+  let e0 = 0;
+  let e1 = 0;
+  for (let i = 0; i < win; i++) {
+    const a = buf[at + i];
+    const b = buf[at + i + lag];
+    r += a * b;
+    e0 += a * a;
+    e1 += b * b;
+  }
+  const denom = Math.sqrt(e0 * e1);
+  return denom > 1e-12 ? r / denom : 0;
+}
+function analyze(wav) {
+  const mono = resample(toMono(wav), RATE);
+  const x = mono.samples;
+  const n = Math.max(1, Math.floor((x.length - FFT_SIZE) / HOP) + 1);
+  const rmsDb = new Float32Array(n);
+  const highRatio = new Float32Array(n);
+  const highDb = new Float32Array(n);
+  const flux = new Float32Array(n);
+  const voiced = new Float32Array(n);
+  const window = new Float32Array(FFT_SIZE);
+  for (let i = 0; i < FFT_SIZE; i++) {
+    window[i] = 0.5 - 0.5 * Math.cos(2 * Math.PI * i / FFT_SIZE);
+  }
+  const re = new Float32Array(FFT_SIZE);
+  const im = new Float32Array(FFT_SIZE);
+  const mag = new Float32Array(BINS);
+  const prevMag = new Float32Array(BINS);
+  const highBin = Math.floor(4e3 / (RATE / 2) * BINS);
+  for (let t = 0; t < n; t++) {
+    const off = t * HOP;
+    let energy = 0;
+    for (let i = 0; i < FFT_SIZE; i++) {
+      const s = x[off + i] ?? 0;
+      energy += s * s;
+      re[i] = s * window[i];
+      im[i] = 0;
+    }
+    rmsDb[t] = 10 * Math.log10(energy / FFT_SIZE + 1e-12);
+    fft(re, im);
+    let lowSum = 0;
+    let highSum = 0;
+    let fluxSum = 0;
+    for (let b = 1; b < BINS; b++) {
+      const m = Math.sqrt(re[b] * re[b] + im[b] * im[b]);
+      mag[b] = m;
+      if (b < highBin) lowSum += m;
+      else highSum += m;
+      const d = Math.log10(m + 1e-6) - Math.log10(prevMag[b] + 1e-6);
+      if (d > 0) fluxSum += d;
+    }
+    prevMag.set(mag);
+    highRatio[t] = highSum / (lowSum + highSum + 1e-12);
+    highDb[t] = 20 * Math.log10(highSum / BINS + 1e-9);
+    flux[t] = t === 0 ? 0 : fluxSum;
+  }
+  const smoothDb = smooth(rmsDb, Math.round(30 / HOP_MS));
+  const smoothHighDb = smooth(highDb, Math.round(20 / HOP_MS));
+  const highFloorDb = quietestWindow(smoothHighDb);
+  const floorDb = percentile(smoothDb, 0.1);
+  const quietDb = quietestWindow(smoothDb);
+  let peakDb = -120;
+  for (let t = 0; t < n; t++) if (smoothDb[t] > peakDb) peakDb = smoothDb[t];
+  const pitchBuf = new Float32Array(Math.floor(x.length / PITCH_DECIM));
+  for (let i = 0; i < pitchBuf.length; i++) {
+    pitchBuf[i] = (x[i * PITCH_DECIM] + x[i * PITCH_DECIM + 1]) * 0.5;
+  }
+  const minLag = Math.floor(PITCH_RATE / MAX_F0);
+  const maxLag = Math.floor(PITCH_RATE / MIN_F0);
+  const loud = [];
+  for (let t = 0; t < n; t++) {
+    if (smoothDb[t] > peakDb - 10) loud.push(t);
+  }
+  const lags = [];
+  const step = Math.max(1, Math.floor(loud.length / 24));
+  for (let k = 0; k < loud.length; k += step) {
+    const at = Math.floor(loud[k] * HOP / PITCH_DECIM);
+    if (at + PITCH_WIN + maxLag >= pitchBuf.length) continue;
+    let best = 0;
+    let bestLag = 0;
+    for (let lag = minLag; lag <= maxLag; lag++) {
+      const v = ncc(pitchBuf, at, PITCH_WIN, lag);
+      if (v > best) {
+        best = v;
+        bestLag = lag;
+      }
+    }
+    if (best > 0.5 && bestLag > 0) lags.push(bestLag);
+  }
+  lags.sort((a, b) => a - b);
+  const centreLag = lags.length ? lags[lags.length >> 1] : 0;
+  const f0 = centreLag ? PITCH_RATE / centreLag : 0;
+  if (centreLag) {
+    const lo = Math.max(minLag, Math.floor(centreLag / 1.7));
+    const hi = Math.min(maxLag, Math.ceil(centreLag * 1.7));
+    for (let t = 0; t < n; t++) {
+      if (smoothDb[t] < floorDb + 6) continue;
+      const at = Math.floor(t * HOP / PITCH_DECIM);
+      if (at + PITCH_WIN + hi >= pitchBuf.length) continue;
+      let best = 0;
+      for (let lag = lo; lag <= hi; lag++) {
+        const v = ncc(pitchBuf, at, PITCH_WIN, lag);
+        if (v > best) best = v;
+      }
+      voiced[t] = best;
+    }
+  }
+  const medianFlux = percentile(flux, 0.5) || 1;
+  for (let t = 0; t < n; t++) flux[t] /= medianFlux;
+  return {
+    n,
+    durationMs: x.length / RATE * 1e3,
+    rmsDb,
+    smoothDb,
+    voiced,
+    highRatio,
+    highDb: smoothHighDb,
+    highFloorDb,
+    flux: smooth(flux, 3),
+    f0,
+    floorDb,
+    quietDb,
+    peakDb
+  };
+}
+function smooth(src, width) {
+  if (width <= 1) return Float32Array.from(src);
+  const prefix = new Float64Array(src.length + 1);
+  for (let i = 0; i < src.length; i++) prefix[i + 1] = prefix[i] + src[i];
+  const out = new Float32Array(src.length);
+  const half = width >> 1;
+  for (let i = 0; i < src.length; i++) {
+    const lo = Math.max(0, i - half);
+    const hi = Math.min(src.length, i + half + 1);
+    out[i] = (prefix[hi] - prefix[lo]) / (hi - lo);
+  }
+  return out;
+}
+function analyzeWav(buf) {
+  return analyze(parseWav(buf));
+}
+
+// src/oto/estimate.ts
+var ARTICULATION = {
+  vowel: {
+    maxConsonantMs: 40,
+    leadInMs: 3,
+    overlapRatio: 0,
+    fixedOverlapMs: 20,
+    voicedConsonant: true,
+    fricationOnset: false
+  },
+  nasalN: {
+    maxConsonantMs: 40,
+    leadInMs: 3,
+    overlapRatio: 0,
+    fixedOverlapMs: 20,
+    voicedConsonant: true,
+    fricationOnset: false
+  },
+  nasal: {
+    maxConsonantMs: 110,
+    leadInMs: 5,
+    overlapRatio: 0.6,
+    voicedConsonant: true,
+    fricationOnset: false
+  },
+  liquid: {
+    maxConsonantMs: 80,
+    leadInMs: 5,
+    overlapRatio: 0.6,
+    voicedConsonant: true,
+    fricationOnset: false
+  },
+  semivowel: {
+    maxConsonantMs: 75,
+    leadInMs: 5,
+    overlapRatio: 0.6,
+    voicedConsonant: true,
+    fricationOnset: false
+  },
+  fricativeVoiceless: {
+    maxConsonantMs: 140,
+    leadInMs: 5,
+    overlapRatio: 0.3,
+    voicedConsonant: false,
+    fricationOnset: true
+  },
+  fricativeVoiced: {
+    maxConsonantMs: 130,
+    leadInMs: 5,
+    overlapRatio: 0.4,
+    voicedConsonant: true,
+    fricationOnset: true
+  },
+  affricate: {
+    maxConsonantMs: 120,
+    leadInMs: 5,
+    overlapRatio: 0.3,
+    voicedConsonant: false,
+    fricationOnset: true
+  },
+  plosiveVoiceless: {
+    maxConsonantMs: 90,
+    leadInMs: 8,
+    overlapRatio: 0,
+    // か/た/ぱ行 need a gap, not a crossfade: the wiki asks for a negative
+    // overlap so the closure that precedes the burst survives synthesis.
+    fixedOverlapMs: -10,
+    voicedConsonant: false,
+    fricationOnset: false
+  },
+  plosiveVoiced: {
+    maxConsonantMs: 90,
+    leadInMs: 8,
+    overlapRatio: 0.55,
+    voicedConsonant: true,
+    fricationOnset: false
+  }
+};
+var MIN_OVERLAP_MS = 12;
+var MAX_OVERLAP_MS = 40;
+var VOICED_THRESHOLD = 0.55;
+var VOICED_RUN_MS = 24;
+var MIN_VOICED_CONSONANT_MS = 22;
+var SEQUENCE_ONSET_BIAS = 12;
+var MAX_SEQUENCE_PRE_MS = 250;
+var DECAY_DROP_DB = 4;
+var RELEASE_GUARD_FRAMES = 10;
+function clamp2(v, lo, hi) {
+  return Math.min(hi, Math.max(lo, v));
+}
+var SOUND_OVER_FLOOR_DB = 8;
+var SOUND_UNDER_PEAK_DB = 15;
+var ATTACK_SLOPE_DB = 1;
+var FRICATION_HIGH_RATIO = 0.45;
+var FRICATION_UNDER_PEAK_DB = 30;
+function soundThreshold(f) {
+  return Math.min(
+    Math.max(f.floorDb + SOUND_OVER_FLOOR_DB, f.peakDb - 34),
+    f.peakDb - SOUND_UNDER_PEAK_DB
+  );
+}
+function findSoundStart(f, from, to) {
+  const th = soundThreshold(f);
+  for (let t = Math.max(0, from); t < Math.min(f.n, to); t++) {
+    if (f.smoothDb[t] >= th) return t;
+  }
+  return Math.max(0, from);
+}
+function findSoundEnd(f, from, to) {
+  const th = soundThreshold(f);
+  for (let t = Math.min(f.n, to) - 1; t > from; t--) {
+    if (f.smoothDb[t] >= th) return t;
+  }
+  return Math.min(f.n - 1, to);
+}
+function findConsonantStart(f, from, to, frication) {
+  const lo = Math.max(0, from);
+  let start = findSoundStart(f, from, to);
+  if (frication) {
+    const level = f.peakDb - FRICATION_UNDER_PEAK_DB;
+    const absolute = f.highFloorDb + 8;
+    for (let t = lo; t < start; t++) {
+      const byRatio = f.highRatio[t] >= FRICATION_HIGH_RATIO && f.smoothDb[t] >= level;
+      if (byRatio || f.highDb[t] >= absolute) {
+        start = t;
+        break;
+      }
+    }
+  }
+  const foot = f.quietDb + 6;
+  const k = Math.max(1, msToFrame(6));
+  while (start > lo) {
+    const back = Math.max(lo, start - k);
+    const stillFalling = f.smoothDb[start] - f.smoothDb[back] >= ATTACK_SLOPE_DB;
+    if (!stillFalling && f.smoothDb[start - 1] <= foot) break;
+    start--;
+  }
+  return start;
+}
+function findVoiceOnset(f, from, to) {
+  const need = Math.ceil(VOICED_RUN_MS / HOP_MS);
+  const hi = Math.min(f.n, to);
+  for (let t = Math.max(0, from); t < hi; t++) {
+    if (f.voiced[t] < VOICED_THRESHOLD) continue;
+    let run = 0;
+    while (t + run < hi && f.voiced[t + run] >= VOICED_THRESHOLD) run++;
+    if (run >= need) return t;
+    t += run;
+  }
+  return -1;
+}
+function findVoicedRelease(f, from, to) {
+  const lo = Math.max(1, from);
+  const hi = Math.min(f.n - 1, to);
+  if (hi <= lo) return lo;
+  const k = Math.max(1, msToFrame(10));
+  let best = lo;
+  let bestScore = -Infinity;
+  for (let t = lo; t < hi; t++) {
+    const rise = f.smoothDb[Math.min(f.n - 1, t + k)] - f.smoothDb[Math.max(0, t - k)];
+    const deSibilance = (f.highRatio[Math.max(0, t - k)] - f.highRatio[Math.min(f.n - 1, t + k)]) * 30;
+    const score = rise + deSibilance + f.flux[t];
+    if (score > bestScore) {
+      bestScore = score;
+      best = t;
+    }
+  }
+  return best;
+}
+function findVowelStable(f, vowelOnset, limit) {
+  const minMs = 45;
+  const maxMs = 110;
+  const from = vowelOnset + msToFrame(minMs);
+  const to = Math.min(limit, vowelOnset + msToFrame(maxMs));
+  if (to <= from) return Math.min(limit, vowelOnset + msToFrame(minMs));
+  const calm = smooth(f.flux, Math.max(1, msToFrame(12)));
+  const need = Math.ceil(20 / HOP_MS);
+  for (let t = from; t < to - need; t++) {
+    let steady = true;
+    for (let k = 0; k < need; k++) {
+      if (calm[t + k] > 1.05) {
+        steady = false;
+        break;
+      }
+    }
+    if (steady) return t;
+  }
+  return to;
+}
+function findDecayStart(f, from, soundEnd) {
+  let peak = from;
+  for (let t = from; t <= soundEnd; t++) {
+    if (f.smoothDb[t] > f.smoothDb[peak]) peak = t;
+  }
+  const th = f.smoothDb[peak] - DECAY_DROP_DB;
+  const need = Math.ceil(24 / HOP_MS);
+  for (let t = peak; t <= soundEnd - need; t++) {
+    let falling = true;
+    for (let k = 0; k < need; k++) {
+      if (f.smoothDb[t + k] >= th) {
+        falling = false;
+        break;
+      }
+    }
+    if (falling) return Math.max(from, t - RELEASE_GUARD_FRAMES);
+  }
+  return Math.max(from, soundEnd - RELEASE_GUARD_FRAMES);
+}
+function hasPrevoicing(f, consStart, vowelOnset) {
+  if (vowelOnset - consStart < msToFrame(25)) return false;
+  let voicedFrames = 0;
+  for (let t = consStart; t < vowelOnset; t++) {
+    if (f.voiced[t] >= VOICED_THRESHOLD) voicedFrames++;
+  }
+  return voicedFrames >= (vowelOnset - consStart) * 0.6;
+}
+function locateMora(f, cls, from, to) {
+  const art = ARTICULATION[cls];
+  const consStart = findConsonantStart(f, from, to, art.fricationOnset);
+  const searchEnd = Math.min(
+    to,
+    consStart + msToFrame(art.maxConsonantMs + 60)
+  );
+  let vowelOnset;
+  if (!art.voicedConsonant) {
+    const v = findVoiceOnset(f, consStart, searchEnd);
+    vowelOnset = v >= 0 ? v : consStart + msToFrame(30);
+  } else if (cls === "vowel" || cls === "nasalN") {
+    const v = findVoiceOnset(f, consStart, searchEnd);
+    vowelOnset = v >= 0 ? v : consStart;
+  } else {
+    const voiceStart = findVoiceOnset(f, consStart, searchEnd);
+    const base = voiceStart >= 0 ? voiceStart : consStart;
+    vowelOnset = findVoicedRelease(
+      f,
+      base + msToFrame(MIN_VOICED_CONSONANT_MS),
+      Math.min(to, base + msToFrame(art.maxConsonantMs))
+    );
+  }
+  vowelOnset = clamp2(vowelOnset, consStart, Math.max(consStart, to - 1));
+  return {
+    consStart,
+    vowelOnset,
+    stable: findVowelStable(f, vowelOnset, to),
+    prevoiced: cls === "plosiveVoiced" && hasPrevoicing(f, consStart, vowelOnset)
+  };
+}
+function round(ms) {
+  return Math.round(ms * 1e3) / 1e3;
+}
+function buildEntry(wav, alias, f, cls, pos, endFrame) {
+  const art = ARTICULATION[cls];
+  let offsetMs = frameTimeMs(pos.consStart) - art.leadInMs;
+  const vowelMs = frameTimeMs(pos.vowelOnset);
+  offsetMs = Math.max(offsetMs, vowelMs - art.maxConsonantMs);
+  offsetMs = clamp2(offsetMs, 0, Math.max(0, vowelMs - 1));
+  const pre = Math.max(1, vowelMs - offsetMs);
+  let overlap;
+  if (art.fixedOverlapMs !== void 0) {
+    overlap = art.fixedOverlapMs;
+  } else {
+    overlap = clamp2(
+      art.overlapRatio * pre,
+      Math.min(MIN_OVERLAP_MS, pre),
+      Math.min(MAX_OVERLAP_MS, pre)
+    );
+  }
+  const consonant = Math.max(
+    pre + 10,
+    Math.abs(overlap) + 10,
+    frameTimeMs(pos.stable) - offsetMs
+  );
+  const endMs = Math.max(frameTimeMs(endFrame), offsetMs + consonant + 30);
+  const cutoff = -(Math.min(endMs, f.durationMs) - offsetMs);
+  return {
+    wav,
+    alias,
+    offset: round(offsetMs),
+    consonant: round(consonant),
+    cutoff: round(cutoff),
+    pre: round(pre),
+    overlap: round(overlap)
+  };
+}
+function estimateSolo(wav, f, syl, aliases) {
+  const soundStart = findSoundStart(f, 0, f.n);
+  const soundEnd = findSoundEnd(f, soundStart, f.n);
+  const pos = locateMora(f, syl.cls, 0, soundEnd);
+  const end = findDecayStart(f, pos.stable, soundEnd);
+  return aliases.map((alias) => buildEntry(wav, alias, f, syl.cls, pos, end));
+}
+function estimateVowelJoin(wav, f, syl, alias) {
+  const soundStart = findSoundStart(f, 0, f.n);
+  const soundEnd = findSoundEnd(f, soundStart, f.n);
+  const pos = locateMora(f, syl.cls, 0, soundEnd);
+  const end = findDecayStart(f, pos.stable, soundEnd);
+  const pre = 50;
+  const overlap = 100;
+  const consonant = 110;
+  const offsetMs = frameTimeMs(pos.stable);
+  const endMs = frameTimeMs(end);
+  if (offsetMs < 0 || endMs - offsetMs < consonant + 40) return null;
+  return {
+    wav,
+    alias,
+    offset: round(offsetMs),
+    consonant,
+    cutoff: round(-(endMs - offsetMs)),
+    pre,
+    overlap
+  };
+}
+var GRID_SNAP_WEIGHT = 0.6;
+function fitLine(values) {
+  const n = values.length;
+  if (n < 2) return { intercept: values[0] ?? 0, slope: 0 };
+  let sx = 0;
+  let sy = 0;
+  let sxx = 0;
+  let sxy = 0;
+  for (let i = 0; i < n; i++) {
+    sx += i;
+    sy += values[i];
+    sxx += i * i;
+    sxy += i * values[i];
+  }
+  const denom = n * sxx - sx * sx;
+  const slope = denom !== 0 ? (n * sxy - sx * sy) / denom : 0;
+  return { intercept: (sy - slope * sx) / n, slope };
+}
+function onsetStrength(f) {
+  const k = Math.max(1, msToFrame(12));
+  const out = new Float32Array(f.n);
+  for (let t = 0; t < f.n; t++) {
+    const rise = f.smoothDb[Math.min(f.n - 1, t + k)] - f.smoothDb[Math.max(0, t - k)];
+    out[t] = f.flux[t] + Math.max(0, rise) / 3;
+  }
+  return smooth(out, Math.max(1, msToFrame(8)));
+}
+function detectGrid(f, count) {
+  if (count < 1) return null;
+  const strength = onsetStrength(f);
+  const uttStart = findSoundStart(f, 0, f.n);
+  const uttEnd = findSoundEnd(f, uttStart, f.n);
+  const span = uttEnd - uttStart;
+  if (span < msToFrame(80)) return null;
+  if (count === 1) {
+    return { start: uttStart, interval: span, onsets: [uttStart] };
+  }
+  const candidates = intervalCandidates(
+    strength,
+    uttStart,
+    uttEnd,
+    span / count
+  );
+  const best = { score: -Infinity, start: uttStart, interval: span / count };
+  const search = (requireFit) => {
+    for (const candidate of candidates) {
+      for (let iv = candidate * 0.85; iv <= candidate * 1.18; iv += candidate * 0.02) {
+        const near2 = windowArgMax(strength, Math.round(iv * 0.22));
+        const tol = iv * 0.22;
+        const hardEnd = uttEnd + iv * 0.4;
+        for (let phase = uttStart - iv * 0.25; phase <= uttStart + iv * 0.5; phase += 1) {
+          if (requireFit && phase + (count - 1) * iv > hardEnd) break;
+          let total = 0;
+          for (let i = 0; i < count; i++) {
+            const centre = phase + i * iv;
+            const at = near2[clamp2(Math.round(centre), 0, f.n - 1)];
+            total += strength[at] - Math.abs(at - centre) / tol;
+          }
+          if (total > best.score) {
+            best.score = total;
+            best.start = phase;
+            best.interval = iv;
+          }
+        }
+      }
+    }
+  };
+  search(true);
+  if (!Number.isFinite(best.score)) search(false);
+  if (!Number.isFinite(best.score)) return null;
+  const near = windowArgMax(strength, Math.round(best.interval * 0.25));
+  const snapped = [];
+  for (let i = 0; i < count; i++) {
+    snapped.push(
+      near[clamp2(Math.round(best.start + i * best.interval), 0, f.n - 1)]
+    );
+  }
+  const { intercept, slope } = fitLine(snapped);
+  const onsets = [];
+  for (let i = 0; i < count; i++) {
+    const fitted = intercept + slope * i;
+    const blended = Math.round(
+      fitted + (snapped[i] - fitted) * GRID_SNAP_WEIGHT
+    );
+    onsets.push(i > 0 ? Math.max(blended, onsets[i - 1] + 4) : blended);
+  }
+  return {
+    start: onsets[0],
+    interval: slope > 0 ? slope : best.interval,
+    onsets
+  };
+}
+function windowArgMax(src, half) {
+  const n = src.length;
+  const out = new Int32Array(n);
+  const deque = new Int32Array(n);
+  let head = 0;
+  let tail = 0;
+  let next = 0;
+  for (let i = 0; i < n; i++) {
+    const limit = Math.min(n - 1, i + half);
+    while (next <= limit) {
+      while (tail > head && src[deque[tail - 1]] <= src[next]) tail--;
+      deque[tail++] = next++;
+    }
+    while (deque[head] < i - half) head++;
+    out[i] = deque[head];
+  }
+  return out;
+}
+function intervalCandidates(strength, from, to, fromSpan) {
+  const lo = msToFrame(180);
+  const hi = Math.min(msToFrame(1300), Math.floor((to - from) / 2));
+  const out = [];
+  if (hi > lo) {
+    let mean = 0;
+    for (let t = from; t < to; t++) mean += strength[t];
+    mean /= Math.max(1, to - from);
+    const r = new Float64Array(hi + 1);
+    let peak = 0;
+    for (let lag = lo; lag <= hi; lag++) {
+      let sum = 0;
+      let n = 0;
+      for (let t = from; t + lag < to; t++, n++) {
+        sum += (strength[t] - mean) * (strength[t + lag] - mean);
+      }
+      r[lag] = n > 0 ? sum / n : 0;
+      if (r[lag] > peak) peak = r[lag];
+    }
+    if (peak > 0) {
+      for (let lag = lo + 1; lag < hi; lag++) {
+        if (r[lag] < peak * 0.7) continue;
+        if (r[lag] < r[lag - 1] || r[lag] < r[lag + 1]) continue;
+        out.push(lag);
+        if (out.length >= 2) break;
+      }
+    }
+  }
+  out.push(fromSpan);
+  const withSubmultiples = [];
+  const seen = /* @__PURE__ */ new Set();
+  const shortest = msToFrame(150);
+  for (const c of out) {
+    for (const divisor of [1, 2, 3]) {
+      const key = Math.round(c / divisor);
+      if (key < shortest || seen.has(key)) continue;
+      seen.add(key);
+      withSubmultiples.push(key);
+    }
+  }
+  if (withSubmultiples.length === 0) withSubmultiples.push(shortest);
+  return withSubmultiples;
+}
+function sequenceVowelOnset(f, cls, onset, next) {
+  if (ARTICULATION[cls].voicedConsonant) return onset + SEQUENCE_ONSET_BIAS;
+  const lo = Math.max(0, onset - msToFrame(100));
+  const hi = Math.min(next, onset + msToFrame(150));
+  let quietest = onset;
+  for (let t = lo; t < hi; t++) {
+    if (f.voiced[t] < f.voiced[quietest]) quietest = t;
+  }
+  if (f.voiced[quietest] >= VOICED_THRESHOLD) return onset;
+  const resumed = findVoiceOnset(
+    f,
+    quietest,
+    Math.min(next, hi + msToFrame(80))
+  );
+  return resumed >= 0 ? resumed : onset;
+}
+function estimateSequence(wav, f, syllables, opts = {}) {
+  const grid = detectGrid(f, syllables.length);
+  if (!grid) return [];
+  const intervalMs = framesToMs(grid.interval);
+  const pre = clamp2(intervalMs / 2, 60, MAX_SEQUENCE_PRE_MS);
+  const overlap = round(pre / 3);
+  const consonant = round(pre * 1.5);
+  const suffix = opts.suffix ?? "";
+  const prefix = opts.prefix ?? "";
+  const entries = [];
+  for (let i = 0; i < syllables.length; i++) {
+    const syl = syllables[i];
+    const onset = grid.onsets[i];
+    const next = i + 1 < syllables.length ? grid.onsets[i + 1] : findSoundEnd(f, onset, f.n);
+    let vowelFrame;
+    if (i === 0) {
+      vowelFrame = locateMora(f, syl.cls, 0, next).vowelOnset;
+    } else {
+      vowelFrame = sequenceVowelOnset(f, syl.cls, onset, next);
+    }
+    const noteMs = frameTimeMs(vowelFrame);
+    const offsetMs = clamp2(noteMs - pre, 0, Math.max(0, noteMs - 1));
+    const actualPre = noteMs - offsetMs;
+    const endMs = Math.min(f.durationMs, noteMs + intervalMs * 2 / 3);
+    const alias = i === 0 ? `- ${prefix}${syl.kana}${suffix}` : `${syllables[i - 1].vowel} ${syl.kana}${suffix}`;
+    entries.push({
+      wav,
+      alias,
+      offset: round(offsetMs),
+      consonant: Math.max(round(consonant), round(actualPre + 20)),
+      cutoff: round(-Math.max(endMs - offsetMs, consonant + 40)),
+      pre: round(actualPre),
+      overlap: i === 0 ? 0 : Math.min(overlap, round(actualPre))
+    });
+  }
+  if (opts.trailingRest && syllables.length > 0) {
+    const last = syllables[syllables.length - 1];
+    const lastOnset = grid.onsets[syllables.length - 1];
+    const soundEnd = findSoundEnd(f, lastOnset, f.n);
+    const decay = findDecayStart(f, lastOnset, soundEnd);
+    const noteMs = frameTimeMs(decay);
+    const offsetMs = clamp2(noteMs - pre, 0, Math.max(0, noteMs - 1));
+    const actualPre = noteMs - offsetMs;
+    entries.push({
+      wav,
+      alias: `${last.vowel} R${suffix}`,
+      offset: round(offsetMs),
+      consonant: round(consonant),
+      cutoff: round(
+        -(Math.min(f.durationMs, frameTimeMs(soundEnd)) - offsetMs)
+      ),
+      pre: round(actualPre),
+      overlap: Math.min(overlap, round(actualPre))
+    });
+  }
+  return entries;
+}
+
+// src/oto/kana.ts
+var CLASS_OF = {
+  "": "vowel",
+  k: "plosiveVoiceless",
+  ky: "plosiveVoiceless",
+  t: "plosiveVoiceless",
+  ty: "plosiveVoiceless",
+  p: "plosiveVoiceless",
+  py: "plosiveVoiceless",
+  g: "plosiveVoiced",
+  gy: "plosiveVoiced",
+  d: "plosiveVoiced",
+  dy: "plosiveVoiced",
+  b: "plosiveVoiced",
+  by: "plosiveVoiced",
+  s: "fricativeVoiceless",
+  sh: "fricativeVoiceless",
+  h: "fricativeVoiceless",
+  hy: "fricativeVoiceless",
+  f: "fricativeVoiceless",
+  z: "fricativeVoiced",
+  j: "fricativeVoiced",
+  v: "fricativeVoiced",
+  ts: "affricate",
+  ch: "affricate",
+  n: "nasal",
+  ny: "nasal",
+  m: "nasal",
+  my: "nasal",
+  r: "liquid",
+  ry: "liquid",
+  y: "semivowel",
+  w: "semivowel"
+};
+var KANA = {
+  \u3042: " a",
+  \u3044: " i",
+  \u3046: " u",
+  \u3048: " e",
+  \u304A: " o",
+  \u3093: " n",
+  \u304B: "k a",
+  \u304D: "k i",
+  \u304F: "k u",
+  \u3051: "k e",
+  \u3053: "k o",
+  \u304C: "g a",
+  \u304E: "g i",
+  \u3050: "g u",
+  \u3052: "g e",
+  \u3054: "g o",
+  \u3055: "s a",
+  \u3057: "sh i",
+  \u3059: "s u",
+  \u305B: "s e",
+  \u305D: "s o",
+  \u3056: "z a",
+  \u3058: "j i",
+  \u305A: "z u",
+  \u305C: "z e",
+  \u305E: "z o",
+  \u305F: "t a",
+  \u3061: "ch i",
+  \u3064: "ts u",
+  \u3066: "t e",
+  \u3068: "t o",
+  \u3060: "d a",
+  \u3062: "j i",
+  \u3065: "z u",
+  \u3067: "d e",
+  \u3069: "d o",
+  \u306A: "n a",
+  \u306B: "n i",
+  \u306C: "n u",
+  \u306D: "n e",
+  \u306E: "n o",
+  \u306F: "h a",
+  \u3072: "h i",
+  \u3075: "f u",
+  \u3078: "h e",
+  \u307B: "h o",
+  \u3070: "b a",
+  \u3073: "b i",
+  \u3076: "b u",
+  \u3079: "b e",
+  \u307C: "b o",
+  \u3071: "p a",
+  \u3074: "p i",
+  \u3077: "p u",
+  \u307A: "p e",
+  \u307D: "p o",
+  \u307E: "m a",
+  \u307F: "m i",
+  \u3080: "m u",
+  \u3081: "m e",
+  \u3082: "m o",
+  \u3084: "y a",
+  \u3086: "y u",
+  \u3088: "y o",
+  \u3089: "r a",
+  \u308A: "r i",
+  \u308B: "r u",
+  \u308C: "r e",
+  \u308D: "r o",
+  \u308F: "w a",
+  \u3090: "w i",
+  \u3091: "w e",
+  \u3092: "w o",
+  \u3094: "v u"
+};
+var SMALL = {
+  \u3041: "a",
+  \u3043: "i",
+  \u3045: "u",
+  \u3047: "e",
+  \u3049: "o",
+  \u3083: "a",
+  \u3085: "u",
+  \u3087: "o",
+  \u308E: "a"
+};
+var DIGRAPH = {
+  \u3057\u3083: "sh a",
+  \u3057\u3085: "sh u",
+  \u3057\u3087: "sh o",
+  \u3057\u3047: "sh e",
+  \u3058\u3083: "j a",
+  \u3058\u3085: "j u",
+  \u3058\u3087: "j o",
+  \u3058\u3047: "j e",
+  \u3061\u3083: "ch a",
+  \u3061\u3085: "ch u",
+  \u3061\u3087: "ch o",
+  \u3061\u3047: "ch e",
+  \u3062\u3083: "j a",
+  \u3062\u3085: "j u",
+  \u3062\u3087: "j o",
+  \u3064\u3041: "ts a",
+  \u3064\u3043: "ts i",
+  \u3064\u3047: "ts e",
+  \u3064\u3049: "ts o",
+  \u3066\u3043: "t i",
+  \u3066\u3083: "ty a",
+  \u3066\u3085: "ty u",
+  \u3066\u3087: "ty o",
+  \u3067\u3043: "d i",
+  \u3067\u3083: "dy a",
+  \u3067\u3085: "dy u",
+  \u3067\u3087: "dy o",
+  \u3068\u3045: "t u",
+  \u3069\u3045: "d u",
+  \u3075\u3041: "f a",
+  \u3075\u3043: "f i",
+  \u3075\u3047: "f e",
+  \u3075\u3049: "f o",
+  \u3075\u3085: "f u",
+  \u3094\u3041: "v a",
+  \u3094\u3043: "v i",
+  \u3094\u3047: "v e",
+  \u3094\u3049: "v o",
+  \u3094\u3085: "v u",
+  \u304F\u3041: "k a",
+  \u304F\u3043: "k i",
+  \u304F\u3047: "k e",
+  \u304F\u3049: "k o",
+  \u3050\u3041: "g a",
+  \u3050\u3043: "g i",
+  \u3050\u3047: "g e",
+  \u3050\u3049: "g o",
+  \u3059\u3043: "s i",
+  \u305A\u3043: "z i"
+};
+function toHiragana(s) {
+  let out = "";
+  for (const ch of s) {
+    const c = ch.codePointAt(0) ?? 0;
+    out += c >= 12449 && c <= 12534 ? String.fromCodePoint(c - 96) : ch;
+  }
+  return out;
+}
+function make(kana, pair) {
+  const sp = pair.indexOf(" ");
+  const consonant = pair.slice(0, sp);
+  const vowel = pair.slice(sp + 1);
+  return { kana, consonant, vowel, cls: CLASS_OF[consonant] ?? "vowel" };
+}
+function readSyllable(src, hira, i) {
+  const two = hira.slice(i, i + 2);
+  if (two.length === 2) {
+    const explicit = DIGRAPH[two];
+    if (explicit) return { syl: make(src.slice(i, i + 2), explicit), len: 2 };
+    const small = SMALL[two[1]];
+    const base = KANA[two[0]];
+    if (small && base) {
+      const sp = base.indexOf(" ");
+      const baseConsonant = base.slice(0, sp);
+      const baseVowel = base.slice(sp + 1);
+      if (baseVowel === "i" && CLASS_OF[`${baseConsonant}y`]) {
+        const pair = `${baseConsonant}y ${small}`;
+        return { syl: make(src.slice(i, i + 2), pair), len: 2 };
+      }
+      if (!baseConsonant) {
+        const glide = baseVowel === "i" ? "y" : baseVowel === "u" ? "w" : "";
+        const syl = make(src.slice(i, i + 2), `${glide} ${small}`);
+        return { syl: { ...syl, cls: "semivowel" }, len: 2 };
+      }
+      return {
+        syl: make(src.slice(i, i + 2), `${baseConsonant} ${small}`),
+        len: 2
+      };
+    }
+  }
+  const one = KANA[hira[i]];
+  if (one) return { syl: make(src.slice(i, i + 1), one), len: 1 };
+  return null;
+}
+function splitKana(text) {
+  const src = text.normalize("NFC");
+  const hira = toHiragana(src);
+  const out = [];
+  for (let i = 0; i < hira.length; ) {
+    const ch = hira[i];
+    if (ch === "\u3063" || ch === "\u30FC") {
+      i++;
+      continue;
+    }
+    const read = readSyllable(src, hira, i);
+    if (!read) return null;
+    out.push(read.syl);
+    i += read.len;
+  }
+  return out.length > 0 ? out : null;
+}
+
+// src/oto/generate.ts
+var REST = /[RrＲｒ]$|息$|吸$/;
+function transcribe(filename) {
+  let body = filename.replace(/\.[^.]+$/, "");
+  body = body.replace(/^[_\-\s]+/, "");
+  body = body.replace(/_[A-G][#b]?-?\d$/i, "");
+  body = body.trim();
+  if (!body) return null;
+  let mark = "";
+  const arrows = /[↑↓→]+$/.exec(body);
+  if (arrows) {
+    mark = arrows[0];
+    body = body.slice(0, -arrows[0].length);
+  }
+  let trailingRest = false;
+  if (REST.test(body)) {
+    trailingRest = true;
+    body = body.replace(REST, "");
+  }
+  const take = /[A-Za-z0-9'][A-Za-z0-9']?$/.exec(body);
+  if (take && body.length > take[0].length) {
+    mark = take[0] + mark;
+    body = body.slice(0, -take[0].length);
+  }
+  let prefix = "";
+  const head = /^[A-Za-z]{1,2}(?=[ぁ-ゟァ-ヿ])/.exec(body);
+  if (head) {
+    prefix = head[0];
+    body = body.slice(head[0].length);
+  }
+  if (!body) return null;
+  const syllables = splitKana(body);
+  if (!syllables) return null;
+  return { syllables, trailingRest, prefix, mark };
+}
+function suffixFromFolderName(folder) {
+  if (/^[A-G][#b]?-?\d$/.test(folder)) return `_${folder}`;
+  const tagged = /_([A-G][#b]?-?\d)(?![0-9A-Za-z])/.exec(folder);
+  return tagged ? `_${tagged[1]}` : "";
+}
+function aliasesFor(name, headAliases) {
+  return headAliases ? [name, `- ${name}`] : [name];
+}
+function generateOtoForFile(file, options = {}) {
+  const suffix = options.suffix ?? "";
+  const headAliases = options.headAliases ?? true;
+  const vowelJoinAliases = options.vowelJoinAliases ?? true;
+  const transcript = transcribe(file.name);
+  if (!transcript) {
+    return {
+      entries: [],
+      skipped: { wav: file.name, reason: "filename is not kana" },
+      style: null
+    };
+  }
+  let frames;
+  try {
+    frames = analyze(parseWav(file.data));
+  } catch (err2) {
+    return {
+      entries: [],
+      skipped: {
+        wav: file.name,
+        reason: err2 instanceof Error ? err2.message : String(err2)
+      },
+      style: null
+    };
+  }
+  if (frames.n < 8) {
+    return {
+      entries: [],
+      skipped: { wav: file.name, reason: "too short to analyse" },
+      style: null
+    };
+  }
+  const { syllables, trailingRest, prefix, mark } = transcript;
+  if (syllables.length === 1 && !trailingRest) {
+    const syl = syllables[0];
+    const name = `${prefix}${syl.kana}${mark}${suffix}`;
+    const entries2 = estimateSolo(
+      file.name,
+      frames,
+      syl,
+      aliasesFor(name, headAliases)
+    );
+    if (vowelJoinAliases && (syl.cls === "vowel" || syl.cls === "nasalN")) {
+      const join = estimateVowelJoin(file.name, frames, syl, `* ${name}`);
+      if (join) entries2.push(join);
+    }
+    return { entries: entries2, skipped: null, style: "solo" };
+  }
+  const entries = estimateSequence(file.name, frames, syllables, {
+    suffix: `${mark}${suffix}`,
+    prefix,
+    trailingRest
+  });
+  if (entries.length === 0) {
+    return {
+      entries: [],
+      skipped: { wav: file.name, reason: "could not segment the phrase" },
+      style: null
+    };
+  }
+  return { entries, skipped: null, style: "sequence" };
+}
+function generateOto(files, options = {}) {
+  const entries = [];
+  const skipped = [];
+  let solo = 0;
+  let sequence = 0;
+  for (const file of files) {
+    const result = generateOtoForFile(file, options);
+    entries.push(...result.entries);
+    if (result.skipped) skipped.push(result.skipped);
+    if (result.style === "solo") solo++;
+    if (result.style === "sequence") sequence++;
+  }
+  return { entries, skipped, style: summarise(solo, sequence) };
+}
+function summarise(solo, sequence) {
+  if (solo && sequence) return "mixed";
+  if (sequence) return "sequence";
+  if (solo) return "solo";
+  return "empty";
+}
+
+// src/oto/write.ts
+var reverseTable = null;
+function shiftJisTable() {
+  if (reverseTable) return reverseTable;
+  const table = /* @__PURE__ */ new Map();
+  const decoder = new TextDecoder("shift_jis", { fatal: false });
+  for (let b = 0; b < 128; b++) {
+    table.set(decoder.decode(new Uint8Array([b])), b);
+  }
+  for (let b = 161; b <= 223; b++) {
+    table.set(decoder.decode(new Uint8Array([b])), b);
+  }
+  const pair = new Uint8Array(2);
+  for (let lead = 129; lead <= 252; lead++) {
+    if (lead > 159 && lead < 224) continue;
+    pair[0] = lead;
+    for (let trail = 64; trail <= 252; trail++) {
+      if (trail === 127) continue;
+      pair[1] = trail;
+      const ch = decoder.decode(pair);
+      if (ch.length !== 1 || ch === "\uFFFD") continue;
+      if (!table.has(ch)) table.set(ch, lead << 8 | trail);
+    }
+  }
+  reverseTable = table;
+  return table;
+}
+function encodeShiftJis(text) {
+  const table = shiftJisTable();
+  const out = [];
+  for (const ch of text) {
+    const code = table.get(ch);
+    if (code === void 0) out.push(63);
+    else if (code > 255) out.push(code >> 8, code & 255);
+    else out.push(code);
+  }
+  return new Uint8Array(out);
+}
+function num(v) {
+  const s = v.toFixed(3);
+  return s.replace(/\.?0+$/, "") || "0";
+}
+function formatOto(entries) {
+  return entries.map(
+    (e) => `${e.wav}=${e.alias},${num(e.offset)},${num(e.consonant)},${num(e.cutoff)},${num(e.pre)},${num(e.overlap)}`
+  ).join("\r\n").concat("\r\n");
+}
+function encodeOto(entries) {
+  return encodeShiftJis(formatOto(entries));
+}
+
+// src/utautts/UtauTTSAdapter.ts
+var UtauTTSAdapter = class {
+  worldline;
+  constructor(worldline) {
+    this.worldline = worldline;
+  }
+  /**
+   * Wasmモジュールを初期化します。
+   */
+  static async initializeWasm(wasmUrl = "utautts.wasm") {
+    if (typeof Go === "undefined") {
+      throw new Error("wasm_exec.js must be loaded before calling initializeWasm.");
+    }
+    const go = new Go();
+    const response = await fetch(wasmUrl);
+    const buffer = await response.arrayBuffer();
+    const result = await WebAssembly.instantiate(buffer, go.importObject);
+    go.run(result.instance);
+    await new Promise((resolve) => setTimeout(resolve, 50));
+    if (typeof utautts_plan !== "function") {
+      throw new Error("utautts_plan failed to initialize in global scope.");
+    }
+  }
+  /**
+   * koe の VoiceBank から、UtauTTS Wasm に渡すための仮想 oto.ini 辞書を構築します。
+   */
+  buildOtoEntriesFromKoe(bank) {
+    const entries = {};
+    const manifest = bank.manifest;
+    for (const [alias, phoneme] of Object.entries(manifest.phonemes)) {
+      entries[alias] = [{
+        Filename: `v_${alias}.wav`,
+        // Wasm内での識別用ダミーファイル名
+        Alias: alias,
+        Offset: 0,
+        // koe は事前にトリミング済みのため常に0
+        Fixed: phoneme.consonant / 48,
+        Preutterance: phoneme.pre / 48,
+        Overlap: phoneme.overlap / 48,
+        SourceGroup: "koe-bank"
+      }];
+    }
+    return entries;
+  }
+  /**
+   * 指定したテキストを合成するための Plan (発話計画) と音声を生成します。
+   */
+  async synthesizeText(bank, text, tone = "C4") {
+    const otoEntries = this.buildOtoEntriesFromKoe(bank);
+    const request = {
+      text,
+      oto_entries: otoEntries,
+      tone,
+      duration_ms: 0
+      // UtauTTS側に自動計算させる
+    };
+    const response = utautts_plan(JSON.stringify(request));
+    if (!response.success) {
+      throw new Error(`UtauTTS Error: ${response.error}`);
+    }
+    const plan = JSON.parse(response.plan);
+    console.log("Generated TTS Plan:", plan);
+    const phraseUnits = [];
+    let currentPosMs = 0;
+    for (const unit of plan.units) {
+      const pcm = await bank.getPcm(unit.alias);
+      if (!pcm) {
+        console.warn(`PCM not found for alias: ${unit.alias}`);
+        continue;
+      }
+      phraseUnits.push({
+        pcm,
+        posMs: currentPosMs,
+        skipMs: unit.offset_ms,
+        lengthMs: unit.duration_ms,
+        fadeInMs: unit.overlap_ms,
+        fadeOutMs: unit.overlap_ms,
+        // 次ノートとのクロスフェードとして設定
+        consonantMs: unit.consonant_ms
+      });
+      currentPosMs += unit.duration_ms;
+    }
+    const baseHz = 261.63;
+    let pitchInput = baseHz;
+    if (plan.pitch_cents && plan.pitch_frame_ms) {
+      const frameMs = plan.pitch_frame_ms;
+      const centsArr = plan.pitch_cents;
+      pitchInput = (tMs, totalMs) => {
+        const frameIndex = Math.floor(tMs / frameMs);
+        if (frameIndex < 0 || frameIndex >= centsArr.length) return baseHz;
+        const centsOffset = centsArr[frameIndex];
+        return baseHz * Math.pow(2, centsOffset / 1200);
+      };
+    }
+    const outPcm = this.worldline.renderPhrase({
+      units: phraseUnits,
+      pitch: pitchInput
+    });
+    if (outPcm) {
+      console.log(`Successfully generated audio: ${outPcm.length} samples.`);
+    } else {
+      console.error("Audio generation failed.");
+    }
+    return outPcm;
+  }
+};
 export {
   KoeEngine,
   MIN_WORLDLINE_SAMPLES,
+  UtauTTSAdapter,
   VoiceBank,
   WORLDLINE_SAMPLE_RATE,
   Worldline,
+  analyze,
+  analyzeWav,
   detectF0,
+  detectGrid,
+  encodeOto,
+  encodeShiftJis,
+  estimateSequence,
+  estimateSolo,
+  estimateVowelJoin,
+  formatOto,
   frqAverageF0InRange,
   frqFileName,
+  generateOto,
+  generateOtoForFile,
   leadInFromEntry,
+  locateMora,
   normalizePcm,
   noteNameToHz,
   otoRegion,
@@ -1398,9 +3291,15 @@ export {
   readWavPcm48k,
   resample,
   samplesToMs,
+  splitKana,
+  suffixFromFolderName,
+  summarise,
+  toHiragana,
   toInt16,
   toMono,
+  transcribe,
   trimToOto,
-  unzipToFileMap
+  unzipToFileMap,
+  zipFiles
 };
 //# sourceMappingURL=index.js.map