Spaces:
Sleeping
Sleeping
File size: 103,416 Bytes
33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 2221103 8b09b87 2221103 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 86371bb 33a89be 86371bb 2221103 86371bb 33a89be 2f01cc6 33a89be 8b09b87 33a89be 2f01cc6 33a89be 2f01cc6 2221103 33a89be 8b09b87 2f01cc6 33a89be 8b09b87 33a89be 8b09b87 33a89be 2221103 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2221103 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 8b09b87 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 2f01cc6 33a89be 0701921 2f01cc6 0701921 2f01cc6 0701921 2f01cc6 0701921 2f01cc6 0701921 33a89be 0701921 2221103 8b09b87 2221103 8b09b87 2221103 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 | # -*- coding: utf-8 -*-
"""
UTAU oto.ini 导出插件
从 TextGrid 提取音素时间边界,生成 UTAU 音源配置文件
一个 wav 文件可包含多条 oto 配置,无需裁剪音频
"""
import os
import json
import glob
import shutil
import logging
from collections import defaultdict
from typing import Any, Dict, List, Optional, Tuple
from .base import ExportPlugin, PluginOption, OptionType
logger = logging.getLogger(__name__)
# ==================== IPA 音素分类 ====================
# 中文辅音(MFA 输出的 IPA 符号)
CHINESE_CONSONANTS = {
# 双唇音
'p', 'pʰ', 'pʲ', 'pʷ', 'b', 'm', 'f',
# 齿龈音
't', 'tʰ', 'tʲ', 'd', 'n', 'l',
# 软腭音
'k', 'kʰ', 'kʷ', 'ɡ', 'g', 'ŋ', 'x', 'h',
# 齿龈-硬腭音(j, q, x)
'tɕ', 'tɕʰ', 'dʑ', 'ɕ', 'ʑ',
# 齿龈塞擦音(z, c, s)
'ts', 'tsʰ', 'dz', 's', 'z',
# 卷舌音(zh, ch, sh, r)
'ʈʂ', 'ʈʂʰ', 'ɖʐ', 'ʂ', 'ʐ',
# 鼻音和近音
'ɲ', 'j', 'w', 'ɥ',
# 喉塞音
'ʔ',
}
# 中文元音(可能带声调标记)
# 注意:MFA 输出的元音通常是单个音素,复合韵母会被拆分成多个音素
CHINESE_VOWELS = {
# 基本单元音
'a', 'o', 'e', 'i', 'u', 'y', 'ü',
'ə', 'ɛ', 'ɔ', 'ɤ', 'ɨ', 'ʅ', 'ʉ',
# MFA 输出的特殊格式
'aw', 'ej', 'ow', # 双元音的 MFA 表示(ai, ei, ou)
# 舌尖元音(zi, ci, si, zhi, chi, shi, ri)
'z̩', 'ʐ̩',
# 卷舌近音(er)
'ɻ',
# 儿化音
'ɚ',
}
# 中文介音(声母和韵母之间的过渡音)
CHINESE_MEDIALS = {
'j', 'w', 'ɥ', # i, u, ü 介音
}
# 中文韵尾(鼻音和元音韵尾)
CHINESE_CODAS = {
'n', 'ŋ', # 鼻音韵尾
'i', 'u', # 元音韵尾(在复韵母中)
}
# 日语辅音
JAPANESE_CONSONANTS = {
'p', 'b', 'm', 'ɸ',
't', 'd', 'n', 's', 'z', 'ɾ', 'r',
'k', 'ɡ', 'g', 'ŋ', 'h',
'tɕ', 'dʑ', 'ɕ', 'ʑ',
'ts', 'dz',
'ɲ', 'j', 'w',
# 长辅音
'nː', 'sː', 'tː', 'kː', 'pː',
}
# 日语元音
JAPANESE_VOWELS = {
'a', 'i', 'ɯ', 'u', 'e', 'o',
'aː', 'iː', 'ɯː', 'uː', 'eː', 'oː',
}
# 跳过的标记
SKIP_MARKS = {'', 'SP', 'AP', '<unk>', 'spn', 'sil'}
# ==================== 模糊拼字近似音素对照表 ====================
# 声母近似组(同组内音素互为替代,按优先级排序)
FUZZY_CONSONANT_GROUPS = [
('sh', 's'), # 翘舌/平舌
('zh', 'z'), # 翘舌/平舌
('ch', 'c'), # 翘舌/平舌
('l', 'n', 'r'), # 边音/鼻音/卷舌
('f', 'h'), # 唇齿/喉音
]
# 韵母近似组(同组内音素互为替代,按优先级排序)
FUZZY_VOWEL_GROUPS = [
('an', 'ang'), # 前鼻/后鼻
('en', 'eng', 'ong'), # 前鼻/后鼻/后鼻圆唇
('in', 'ing'), # 前鼻/后鼻
('ian', 'iang'), # 前鼻/后鼻
('uan', 'uang'), # 前鼻/后鼻
# i 行韵母近似组(带鼻音韵尾的可以用不带鼻音韵尾的替代)
('ia', 'ian'), # ia ←→ ian(如 xia ←→ xian)
('ie', 'ian'), # ie ←→ ian(如 jie ←→ jian)
('iao', 'ian'), # iao ←→ ian(如 qiao ←→ qian)
('iu', 'in'), # iu ←→ in(如 liu ←→ lin)
# u 行韵母近似组
('ua', 'uan'), # ua ←→ uan(如 kua ←→ kuan)
('uo', 'un'), # uo ←→ un(如 duo ←→ dun)
('ui', 'un'), # ui ←→ un(如 dui ←→ dun)
('uai', 'uan'), # uai ←→ uan(如 kuai ←→ kuan)
# 单元音与复韵母近似组
('a', 'ai', 'ao', 'an'), # a 系列
('o', 'ou', 'ong'), # o 系列
('e', 'ei', 'en'), # e 系列
]
def is_consonant(phone: str, language: str) -> bool:
"""判断音素是否为辅音"""
base_phone = _strip_tone(phone)
if language in ('chinese', 'zh', 'mandarin'):
return base_phone in CHINESE_CONSONANTS
elif language in ('japanese', 'ja', 'jp'):
return base_phone in JAPANESE_CONSONANTS
return False
def is_vowel(phone: str, language: str) -> bool:
"""判断音素是否为元音"""
base_phone = _strip_tone(phone)
if language in ('chinese', 'zh', 'mandarin'):
# 直接匹配
if base_phone in CHINESE_VOWELS:
return True
# 检查是否以元音字符开头(处理复合元音)
vowel_starts = ['a', 'o', 'e', 'i', 'u', 'y', 'ə', 'ɛ', 'ɔ', 'ɤ', 'ɨ', 'ʅ', 'ʉ', 'ɚ']
for v in vowel_starts:
if base_phone.startswith(v):
return True
# 检查特殊的舌尖元音(带组合字符)
if 'z̩' in base_phone or 'ʐ̩' in base_phone:
return True
# 检查卷舌近音
if 'ɻ' in base_phone:
return True
return False
elif language in ('japanese', 'ja', 'jp'):
return base_phone in JAPANESE_VOWELS or base_phone.rstrip('ː') in {'a', 'i', 'ɯ', 'u', 'e', 'o'}
return False
def _strip_tone(phone: str) -> str:
"""移除声调标记"""
tone_marks = '˥˦˧˨˩ˇˊˋ¯'
result = phone
for mark in tone_marks:
result = result.replace(mark, '')
return result
# ==================== IPA 到别名转换 ====================
# 中文 IPA 辅音到拼音声母映射
CHINESE_CONSONANT_TO_PINYIN = {
'p': 'b', 'pʰ': 'p', 'pʲ': 'p', 'pʷ': 'b',
'm': 'm', 'f': 'f',
't': 'd', 'tʰ': 't', 'tʲ': 'd',
'n': 'n', 'l': 'l',
'k': 'g', 'kʰ': 'k', 'kʷ': 'g',
'ɡ': 'g', 'g': 'g',
'x': 'h', 'h': 'h',
'tɕ': 'j', 'tɕʰ': 'q', 'ɕ': 'x',
'ts': 'z', 'tsʰ': 'c', 's': 's',
'ʈʂ': 'zh', 'ʈʂʰ': 'ch', 'ʂ': 'sh', 'ʐ': 'r',
'ɲ': 'n', 'ŋ': '', # ng 不作为声母
'j': '', 'w': '', 'ɥ': '', # 介音不作为声母
'ʔ': '',
}
# 中文 IPA 元音到拼音韵母映射
CHINESE_VOWEL_TO_PINYIN = {
# 单元音韵母
'a': 'a', 'o': 'o', 'e': 'e', 'i': 'i', 'u': 'u', 'y': 'v', 'ü': 'v',
'ə': 'e', 'ɛ': 'e', 'ɔ': 'o', 'ɤ': 'e', 'ɨ': 'i',
# 复韵母(MFA 可能的 IPA 格式)
'aj': 'ai', 'aw': 'ao', 'ej': 'ei', 'ow': 'ou',
'ai': 'ai', 'ao': 'ao', 'ei': 'ei', 'ou': 'ou', # 直接形式
# i 行韵母(MFA 可能的组合形式)
'ja': 'ia', 'je': 'ie', 'jɛ': 'ie', 'jao': 'iao', 'jow': 'iu', 'ju': 'iu',
'ia': 'ia', 'ie': 'ie', 'iao': 'iao', 'iu': 'iu', # 直接形式
# u 行韵母(MFA 可能的组合形式)
'wa': 'ua', 'wo': 'uo', 'wɔ': 'uo', 'wej': 'ui', 'waj': 'uai',
'ua': 'ua', 'uo': 'uo', 'ui': 'ui', 'uai': 'uai', # 直接形式
# ü 行韵母(MFA 可能的组合形式)
'ɥe': 've', 'ɥɛ': 've',
've': 've', 'yue': 've', # 直接形式
# 鼻音韵母(MFA 可能的组合形式)
'an': 'an', 'en': 'en', 'ang': 'ang', 'eng': 'eng', 'ong': 'ong',
'in': 'in', 'ing': 'ing', 'ian': 'ian', 'iang': 'iang', 'iong': 'iong',
'uan': 'uan', 'un': 'un', 'uang': 'uang', 'ueng': 'ueng',
'van': 'van', 'vn': 'vn',
# 舌尖元音
'z̩': 'i', 'ʐ̩': 'i', 'ʅ': 'i',
# 卷舌音
'ɻ': 'er', 'ɚ': 'er',
}
# 介音+元音组合到韵母的映射
MEDIAL_VOWEL_TO_FINAL = {
# j 介音(i 行韵母)
('j', 'a'): 'ia', ('j', 'e'): 'ie', ('j', 'ɛ'): 'ie',
('j', 'aw'): 'iao', ('j', 'o'): 'io',
('j', 'u'): 'iu', ('j', 'ow'): 'iou',
# w 介音(u 行韵母)
('w', 'a'): 'ua', ('w', 'o'): 'uo', ('w', 'ɔ'): 'uo',
('w', 'ej'): 'uei', ('w', 'e'): 'ue',
('w', 'aj'): 'uai', ('w', 'ai'): 'uai',
# ɥ 介音(ü 行韵母)
('ɥ', 'e'): 've', ('ɥ', 'ɛ'): 've',
}
# 介音+元音+韵尾组合到韵母的映射
MEDIAL_VOWEL_CODA_TO_FINAL = {
# j 介音 + 元音 + 韵尾
('j', 'a', 'n'): 'ian', ('j', 'e', 'n'): 'in',
('j', 'a', 'ŋ'): 'iang', ('j', 'o', 'ŋ'): 'iong',
# w 介音 + 元音 + 韵尾
('w', 'a', 'n'): 'uan', ('w', 'ə', 'n'): 'uen', ('w', 'e', 'n'): 'uen',
('w', 'a', 'ŋ'): 'uang', ('w', 'ə', 'ŋ'): 'ueng', ('w', 'e', 'ŋ'): 'ueng',
# ɥ 介音 + 元音 + 韵尾
('ɥ', 'a', 'n'): 'van', ('ɥ', 'e', 'n'): 'vn',
}
# 元音+韵尾组合到拼音韵母的映射
VOWEL_CODA_TO_PINYIN = {
# 前鼻音韵母
('a', 'n'): 'an', ('ə', 'n'): 'en', ('e', 'n'): 'en',
('i', 'n'): 'in', ('y', 'n'): 'un', ('u', 'n'): 'un',
# 后鼻音韵母
('a', 'ŋ'): 'ang', ('ə', 'ŋ'): 'eng', ('e', 'ŋ'): 'eng',
('i', 'ŋ'): 'ing', ('o', 'ŋ'): 'ong', ('u', 'ŋ'): 'ong',
# 复韵母(元音+元音)
('a', 'i'): 'ai', ('e', 'i'): 'ei', ('ej', 'i'): 'ei',
('a', 'u'): 'ao', ('aw', 'u'): 'ao', ('o', 'u'): 'ou', ('ow', 'u'): 'ou',
# i 行韵母
('i', 'a'): 'ia', ('i', 'e'): 'ie', ('i', 'ɛ'): 'ie',
('i', 'u'): 'iu',
# u 行韵母
('u', 'a'): 'ua', ('u', 'o'): 'uo', ('u', 'ɔ'): 'uo',
('u', 'i'): 'ui', ('u', 'e'): 'ue',
# ü 行韵母
('y', 'e'): 've', ('y', 'ɛ'): 've',
}
# IPA 音节组合到标准拼音的映射表(处理特殊组合规则)
IPA_SYLLABLE_TO_PINYIN = {
# j/q/x + ü 系列(ü 简写为 u)
('tɕ', 'y'): 'ju', ('tɕʰ', 'y'): 'qu', ('ɕ', 'y'): 'xu',
('tɕ', 'ɥ'): 'ju', ('tɕʰ', 'ɥ'): 'qu', ('ɕ', 'ɥ'): 'xu',
('tɕ', 'yɛ'): 'jue', ('tɕʰ', 'yɛ'): 'que', ('ɕ', 'yɛ'): 'xue',
('tɕ', 'yan'): 'juan', ('tɕʰ', 'yan'): 'quan', ('ɕ', 'yan'): 'xuan',
('tɕ', 'yn'): 'jun', ('tɕʰ', 'yn'): 'qun', ('ɕ', 'yn'): 'xun',
# 零声母 + i/u/ü 开头的韵母(需要加 y/w)
('', 'i'): 'yi', ('', 'in'): 'yin', ('', 'ing'): 'ying',
('', 'u'): 'wu', ('', 'un'): 'wen', ('', 'ong'): 'weng',
('', 'y'): 'yu', ('', 'yn'): 'yun',
# i 行韵母(ia, ie, iao, ian, iang, iong, iu)
('', 'ia'): 'ya', ('', 'iɛ'): 'ye', ('', 'ie'): 'ye',
('', 'iao'): 'yao', ('', 'ian'): 'yan', ('', 'iang'): 'yang',
('', 'iou'): 'you', ('', 'iu'): 'you',
('', 'iong'): 'yong',
# u 行韵母(ua, uo, uai, uei, uan, uen, uang, ueng)
('', 'ua'): 'wa', ('', 'uɔ'): 'wo', ('', 'uo'): 'wo',
('', 'uai'): 'wai', ('', 'uei'): 'wei', ('', 'ui'): 'wei',
('', 'uan'): 'wan', ('', 'uen'): 'wen',
('', 'uang'): 'wang', ('', 'ueng'): 'weng',
# ü 行韵母(üe, üan, ün)
('', 'yɛ'): 'yue', ('', 'üe'): 'yue',
('', 'yan'): 'yuan', ('', 'üan'): 'yuan',
('', 'yn'): 'yun', ('', 'ün'): 'yun',
# zh/ch/sh/r + i 实际是舌尖元音
('ʈʂ', 'ʐ̩'): 'zhi', ('ʈʂʰ', 'ʐ̩'): 'chi', ('ʂ', 'ʐ̩'): 'shi', ('ʐ', 'ʐ̩'): 'ri',
('ʈʂ', 'z̩'): 'zhi', ('ʈʂʰ', 'z̩'): 'chi', ('ʂ', 'z̩'): 'shi', ('ʐ', 'z̩'): 'ri',
('ʈʂ', 'ʅ'): 'zhi', ('ʈʂʰ', 'ʅ'): 'chi', ('ʂ', 'ʅ'): 'shi', ('ʐ', 'ʅ'): 'ri',
# z/c/s + i 实际是舌尖元音
('ts', 'z̩'): 'zi', ('tsʰ', 'z̩'): 'ci', ('s', 'z̩'): 'si',
('ts', 'ʅ'): 'zi', ('tsʰ', 'ʅ'): 'ci', ('s', 'ʅ'): 'si',
# n/l + ü 系列(保持 ü)
('n', 'y'): 'nv', ('l', 'y'): 'lv',
('n', 'yɛ'): 'nve', ('l', 'yɛ'): 'lve',
# 其他特殊组合
('ʔ', 'a'): 'a', ('ʔ', 'o'): 'o', ('ʔ', 'e'): 'e',
('ʔ', 'ai'): 'ai', ('ʔ', 'ei'): 'ei', ('ʔ', 'ao'): 'ao', ('ʔ', 'ou'): 'ou',
('ʔ', 'an'): 'an', ('ʔ', 'en'): 'en', ('ʔ', 'ang'): 'ang', ('ʔ', 'eng'): 'eng',
('ʔ', 'ej'): 'ei', ('ʔ', 'aw'): 'ao', ('ʔ', 'ow'): 'ou',
# 儿化音
('', 'ɻ'): 'er', ('', 'ɚ'): 'er',
}
# 日语 IPA 到罗马音映射
JAPANESE_IPA_TO_ROMAJI = {
# 辅音
'p': 'p', 'b': 'b', 'm': 'm', 'ɸ': 'f',
't': 't', 'd': 'd', 'n': 'n', 's': 's', 'z': 'z', 'ɾ': 'r', 'r': 'r',
'k': 'k', 'ɡ': 'g', 'g': 'g', 'h': 'h',
'tɕ': 'ch', 'dʑ': 'j', 'ɕ': 'sh', 'ʑ': 'j',
'ts': 'ts', 'dz': 'z',
'ɲ': 'ny', 'ŋ': 'ng', 'j': 'y', 'w': 'w',
# 长辅音(促音后)
'nː': 'n', 'sː': 's', 'tː': 't', 'kː': 'k', 'pː': 'p',
# 元音
'a': 'a', 'i': 'i', 'ɯ': 'u', 'u': 'u', 'e': 'e', 'o': 'o',
'aː': 'a', 'iː': 'i', 'ɯː': 'u', 'uː': 'u', 'eː': 'e', 'oː': 'o',
}
# 罗马音到平假名映射
ROMAJI_TO_HIRAGANA = {
# 基本元音
'a': 'あ', 'i': 'い', 'u': 'う', 'e': 'え', 'o': 'お',
# か行
'ka': 'か', 'ki': 'き', 'ku': 'く', 'ke': 'け', 'ko': 'こ',
# さ行
'sa': 'さ', 'shi': 'し', 'si': 'し', 'su': 'す', 'se': 'せ', 'so': 'そ',
# た行
'ta': 'た', 'chi': 'ち', 'ti': 'ち', 'tsu': 'つ', 'tu': 'つ', 'te': 'て', 'to': 'と',
# な行
'na': 'な', 'ni': 'に', 'nu': 'ぬ', 'ne': 'ね', 'no': 'の',
# は行
'ha': 'は', 'hi': 'ひ', 'fu': 'ふ', 'hu': 'ふ', 'he': 'へ', 'ho': 'ほ',
# ま行
'ma': 'ま', 'mi': 'み', 'mu': 'む', 'me': 'め', 'mo': 'も',
# や行
'ya': 'や', 'yu': 'ゆ', 'yo': 'よ',
# ら行
'ra': 'ら', 'ri': 'り', 'ru': 'る', 're': 'れ', 'ro': 'ろ',
# わ行
'wa': 'わ', 'wo': 'を', 'n': 'ん',
# が行
'ga': 'が', 'gi': 'ぎ', 'gu': 'ぐ', 'ge': 'げ', 'go': 'ご',
# ざ行
'za': 'ざ', 'ji': 'じ', 'zi': 'じ', 'zu': 'ず', 'ze': 'ぜ', 'zo': 'ぞ',
# だ行
'da': 'だ', 'di': 'ぢ', 'du': 'づ', 'de': 'で', 'do': 'ど',
# ば行
'ba': 'ば', 'bi': 'び', 'bu': 'ぶ', 'be': 'べ', 'bo': 'ぼ',
# ぱ行
'pa': 'ぱ', 'pi': 'ぴ', 'pu': 'ぷ', 'pe': 'ぺ', 'po': 'ぽ',
# 拗音
'kya': 'きゃ', 'kyu': 'きゅ', 'kyo': 'きょ',
'sha': 'しゃ', 'shu': 'しゅ', 'sho': 'しょ',
'cha': 'ちゃ', 'chu': 'ちゅ', 'cho': 'ちょ',
'nya': 'にゃ', 'nyu': 'にゅ', 'nyo': 'にょ',
'hya': 'ひゃ', 'hyu': 'ひゅ', 'hyo': 'ひょ',
'mya': 'みゃ', 'myu': 'みゅ', 'myo': 'みょ',
'rya': 'りゃ', 'ryu': 'りゅ', 'ryo': 'りょ',
'gya': 'ぎゃ', 'gyu': 'ぎゅ', 'gyo': 'ぎょ',
'ja': 'じゃ', 'ju': 'じゅ', 'jo': 'じょ',
'bya': 'びゃ', 'byu': 'びゅ', 'byo': 'びょ',
'pya': 'ぴゃ', 'pyu': 'ぴゅ', 'pyo': 'ぴょ',
}
def ipa_to_alias(consonant: Optional[str], vowel: Optional[str], language: str, use_hiragana: bool = False) -> Optional[str]:
"""将 IPA 音素转换为别名(标准拼音或罗马音)"""
c_base = _strip_tone(consonant) if consonant else ''
v_base = _strip_tone(vowel) if vowel else ''
if language in ('chinese', 'zh', 'mandarin'):
# 中文:使用完整的音节转换规则
return _ipa_to_pinyin(c_base, v_base)
else:
# 日语
c_alias = JAPANESE_IPA_TO_ROMAJI.get(c_base, c_base)
v_alias = JAPANESE_IPA_TO_ROMAJI.get(v_base, v_base)
romaji = (c_alias or '') + (v_alias or '')
# 清理非 ASCII
romaji = ''.join(c for c in romaji if c.isascii() and (c.isalnum() or c == '_'))
romaji = romaji.lower()
if not romaji:
return None
if use_hiragana:
# 尝试转换为平假名
return ROMAJI_TO_HIRAGANA.get(romaji, romaji)
return romaji
def _ipa_to_pinyin(consonant: str, vowel: str) -> Optional[str]:
"""
将 IPA 辅音+韵母转换为标准汉语拼音
参数:
consonant: IPA 辅音(已去除声调),可以是空字符串表示零声母
vowel: IPA 韵母(已去除声调),可能是单个元音或元音+韵尾的组合
返回:
标准拼音,如果无法转换则返回 None
"""
# 1. 先查找特殊组合映射
syllable_key = (consonant, vowel)
if syllable_key in IPA_SYLLABLE_TO_PINYIN:
return IPA_SYLLABLE_TO_PINYIN[syllable_key]
# 2. 获取声母的拼音
c_pinyin = ''
if consonant and consonant != 'ʔ':
if consonant in CHINESE_CONSONANT_TO_PINYIN:
c_pinyin = CHINESE_CONSONANT_TO_PINYIN[consonant]
else:
# 未知辅音,无法转换
return None
# 3. 获取韵母的拼音
# 韵母可能是单个元音,也可能是元音+韵尾的组合字符串
v_pinyin = ''
if vowel:
# 直接查找完整韵母
if vowel in CHINESE_VOWEL_TO_PINYIN:
v_pinyin = CHINESE_VOWEL_TO_PINYIN[vowel]
else:
# 韵母可能是组合形式,无法直接映射
# 这种情况应该在 _syllable_to_pinyin 中处理
return None
if not v_pinyin:
return None
# 4. 处理零声母(无声母或喉塞音)
if not c_pinyin:
# 零声母需要根据韵母添加 y/w/yu
if v_pinyin == 'i':
return 'yi'
elif v_pinyin in ('in', 'ing'):
return 'y' + v_pinyin
elif v_pinyin.startswith('i') and len(v_pinyin) > 1:
# ia->ya, ie->ye, iao->yao, ian->yan, iang->yang, iu->you, iong->yong
return 'y' + v_pinyin[1:]
elif v_pinyin == 'u':
return 'wu'
elif v_pinyin == 'un':
return 'wen'
elif v_pinyin == 'ong':
return 'weng'
elif v_pinyin.startswith('u') and len(v_pinyin) > 1:
# ua->wa, uo->wo, uai->wai, ui->wei, uan->wan, uang->wang
return 'w' + v_pinyin[1:]
elif v_pinyin == 'v':
# ü 单独出现写作 yu
return 'yu'
elif v_pinyin.startswith('v') and len(v_pinyin) > 1:
# ve->yue, van->yuan, vn->yun
return 'yu' + v_pinyin[1:]
else:
# a, o, e, ai, ei, ao, ou, an, en, ang, eng, er 等
return v_pinyin
# 5. 有声母的情况
# 5.1 j/q/x + ü 系列:ü 写作 u
if c_pinyin in ('j', 'q', 'x'):
if v_pinyin == 'v':
return c_pinyin + 'u'
elif v_pinyin.startswith('v'):
# jve->jue, jvan->juan, jvn->jun
return c_pinyin + 'u' + v_pinyin[1:]
else:
return c_pinyin + v_pinyin
# 5.2 n/l + ü 系列:保持 v(表示 ü)
elif c_pinyin in ('n', 'l'):
# 只有 n/l 才需要区分 u 和 ü
return c_pinyin + v_pinyin
# 5.3 其他声母 + v:v 改写为 u(因为不会产生歧义)
elif v_pinyin == 'v':
return c_pinyin + 'u'
elif v_pinyin.startswith('v'):
return c_pinyin + 'u' + v_pinyin[1:]
# 5.4 普通组合
else:
return c_pinyin + v_pinyin
class UTAUOtoExportPlugin(ExportPlugin):
"""UTAU oto.ini 导出插件"""
name = "UTAU oto.ini 导出"
description = "从 TextGrid 生成 UTAU 音源配置文件,一个 wav 可包含多条配置"
version = "1.2.0"
author = "内置"
def get_options(self) -> List[PluginOption]:
return [
PluginOption(
key="cross_language",
label="跨语种导出",
option_type=OptionType.SWITCH,
default=False,
description="【TODO】启用中跨日或日跨中的音素映射导出"
),
PluginOption(
key="max_samples",
label="每个别名最大样本数",
option_type=OptionType.NUMBER,
default=5,
min_value=1,
max_value=100,
description="同一别名保留的最大条目数"
),
PluginOption(
key="quality_metrics",
label="质量评估维度",
option_type=OptionType.COMBO,
default="duration+rms",
choices=["duration", "duration+rms", "duration+f0", "all"],
description="duration=仅时长, +rms=音量稳定性, +f0=音高稳定性。选择 all 可能耗时较长"
),
PluginOption(
key="naming_rule",
label="别名命名规则",
option_type=OptionType.TEXT,
default="%p%%n%",
description="变量: %p%=拼音/罗马音, %n%=序号。示例: %p%_%n% → ba_1"
),
PluginOption(
key="first_naming_rule",
label="首个样本命名规则",
option_type=OptionType.TEXT,
default="%p%",
description="第0个样本的特殊规则,留空则使用通用规则。示例: %p% → ba"
),
PluginOption(
key="alias_style",
label="别名风格(日语)",
option_type=OptionType.COMBO,
default="hiragana",
choices=["romaji", "hiragana"],
description="日语音源的别名格式:罗马音或平假名"
),
PluginOption(
key="overlap_ratio",
label="Overlap 比例",
option_type=OptionType.NUMBER,
default=0.3,
min_value=0.1,
max_value=0.5,
description="Overlap = Preutterance × 此比例"
),
PluginOption(
key="auto_phoneme_combine",
label="自动拼字",
option_type=OptionType.SWITCH,
default=False,
description="用已有的高质量音素拼接生成缺失的音素组合"
),
PluginOption(
key="crossfade_ms",
label="拼接淡入淡出时长(ms)",
option_type=OptionType.NUMBER,
default=10,
min_value=5,
max_value=50,
description="自动拼字时辅音与元音之间的交叉淡化时长",
visible_when={"auto_phoneme_combine": True}
),
PluginOption(
key="fuzzy_phoneme",
label="模糊拼字",
option_type=OptionType.SWITCH,
default=False,
description="用近似声母/韵母替代缺失音素(如 sh↔s, an↔ang),仅中文有效",
visible_when={"auto_phoneme_combine": True}
),
PluginOption(
key="encoding",
label="文件编码",
option_type=OptionType.COMBO,
default="shift_jis",
choices=["shift_jis", "utf-8", "gbk"],
description="oto.ini 和 character.txt 编码(UTAU 标准为 Shift_JIS)"
),
PluginOption(
key="character_name",
label="角色名称",
option_type=OptionType.TEXT,
default="",
description="character.txt 中的角色名,留空则使用音源名称"
),
PluginOption(
key="cvvc_mode",
label="CVVC 模式",
option_type=OptionType.SWITCH,
default=False,
description="启用 CVVC 模式,额外生成 VC 部(元音到辅音过渡)条目"
),
PluginOption(
key="vc_alias_separator",
label="VC 别名分隔符",
option_type=OptionType.COMBO,
default=" ",
choices=[" ", "_", "-"],
description="VC 部别名中元音和辅音之间的分隔符",
visible_when={"cvvc_mode": True}
),
PluginOption(
key="vc_offset_ratio",
label="VC 偏移比例",
option_type=OptionType.NUMBER,
default=0.5,
min_value=0.3,
max_value=0.8,
description="VC 部开始位置 = 元音结束位置 - 元音时长 × 此比例",
visible_when={"cvvc_mode": True}
),
PluginOption(
key="vc_overlap_ratio",
label="VC Overlap 比例",
option_type=OptionType.NUMBER,
default=0.5,
min_value=0.3,
max_value=0.8,
description="VC 部的 Overlap = Preutterance × 此比例",
visible_when={"cvvc_mode": True}
),
]
def export(
self,
source_name: str,
bank_dir: str,
options: Dict[str, Any]
) -> Tuple[bool, str]:
"""执行 UTAU oto.ini 导出"""
try:
# 使用基类方法加载语言设置
language = self.load_language_from_meta(bank_dir, source_name)
# 获取选项
max_samples = int(options.get("max_samples", 5))
quality_metrics = options.get("quality_metrics", "duration")
naming_rule = options.get("naming_rule", "%p%%n%")
first_naming_rule = options.get("first_naming_rule", "%p%")
alias_style = options.get("alias_style", "romaji")
overlap_ratio = float(options.get("overlap_ratio", 0.3))
encoding = options.get("encoding", "utf-8")
character_name = options.get("character_name", "").strip()
auto_phoneme_combine = options.get("auto_phoneme_combine", False)
crossfade_ms = int(options.get("crossfade_ms", 10))
fuzzy_phoneme = options.get("fuzzy_phoneme", False)
use_hiragana = (alias_style == "hiragana") and language in ('japanese', 'ja', 'jp')
# CVVC 模式选项
cvvc_mode = options.get("cvvc_mode", False)
vc_separator = options.get("vc_alias_separator", " ")
vc_offset_ratio = float(options.get("vc_offset_ratio", 0.5))
vc_overlap_ratio = float(options.get("vc_overlap_ratio", 0.5))
# 使用基类方法解析质量评估维度
enabled_metrics = self.parse_quality_metrics(quality_metrics)
paths = self.get_source_paths(bank_dir, source_name)
export_dir = self.get_export_dir(bank_dir, source_name, "utau_oto")
os.makedirs(export_dir, exist_ok=True)
# 步骤1: 解析 TextGrid 并生成 oto 条目
if cvvc_mode:
self._log("【解析 TextGrid 文件】(CVVC 模式)")
else:
self._log("【解析 TextGrid 文件】")
oto_entries, wav_files = self._parse_textgrids(
paths["slices_dir"],
paths["textgrid_dir"],
language,
use_hiragana,
overlap_ratio,
cvvc_mode=cvvc_mode,
vc_offset_ratio=vc_offset_ratio,
vc_overlap_ratio=vc_overlap_ratio,
vc_separator=vc_separator
)
if not oto_entries:
return False, "未能从 TextGrid 提取有效音素"
self._log(f"提取到 {len(oto_entries)} 条原始 oto 配置")
# 步骤2: 按别名分组并限制数量,添加编号
self._log(f"\n【筛选最佳样本】评估维度: {enabled_metrics}")
filtered_entries, used_wavs = self._filter_by_alias(
oto_entries, max_samples, naming_rule, first_naming_rule,
paths["slices_dir"], enabled_metrics
)
self._log(f"筛选后保留 {len(filtered_entries)} 条配置,涉及 {len(used_wavs)} 个音频文件")
# 步骤2.5: 自动拼字(如果启用)
combined_count = 0
if auto_phoneme_combine:
self._log("\n【自动拼字】")
combined_entries, combined_wavs = self._auto_combine_phonemes(
oto_entries,
filtered_entries,
paths["slices_dir"],
export_dir,
language,
use_hiragana,
overlap_ratio,
crossfade_ms,
first_naming_rule,
fuzzy_phoneme
)
if combined_entries:
filtered_entries.extend(combined_entries)
used_wavs.update(combined_wavs)
combined_count = len(combined_entries)
self._log(f"拼接生成 {combined_count} 条新配置")
# 步骤3: 复制音频文件(自动检测文件名是否需要转拼音)
self._log("\n【复制音频文件】")
copied, filename_map = self._copy_wav_files(
used_wavs, paths["slices_dir"], export_dir, encoding
)
self._log(f"复制了 {copied} 个音频文件")
# 步骤4: 写入 oto.ini
self._log("\n【生成 oto.ini】")
oto_path = os.path.join(export_dir, "oto.ini")
self._write_oto_ini(filtered_entries, oto_path, encoding, filename_map)
self._log(f"写入: {oto_path}")
# 步骤5: 写入 character.txt
self._log("\n【生成 character.txt】")
char_path = os.path.join(export_dir, "character.txt")
# 使用自定义角色名,留空则使用音源名称
final_character_name = character_name if character_name else source_name
self._write_character_txt(final_character_name, char_path, encoding)
self._log(f"写入: {char_path}")
# 统计别名数量
unique_aliases = set(e["alias"] for e in filtered_entries)
result_msg = f"导出完成: {export_dir}\n{len(unique_aliases)} 个别名,{len(filtered_entries)} 条配置,{copied} 个音频"
if combined_count > 0:
result_msg += f"\n(其中 {combined_count} 条为自动拼接生成)"
return True, result_msg
except Exception as e:
logger.error(f"UTAU oto.ini 导出失败: {e}", exc_info=True)
return False, str(e)
def _parse_textgrids(
self,
slices_dir: str,
textgrid_dir: str,
language: str,
use_hiragana: bool,
overlap_ratio: float,
cvvc_mode: bool = False,
vc_offset_ratio: float = 0.5,
vc_overlap_ratio: float = 0.5,
vc_separator: str = " "
) -> Tuple[List[Dict], set]:
"""解析 TextGrid 文件,提取音素边界
参数:
slices_dir: 切片目录
textgrid_dir: TextGrid 目录
language: 语言
use_hiragana: 是否使用平假名
overlap_ratio: CV 部 overlap 比例
cvvc_mode: 是否启用 CVVC 模式
vc_offset_ratio: VC 偏移比例
vc_overlap_ratio: VC overlap 比例
vc_separator: VC 别名分隔符
"""
import textgrid
import soundfile as sf
tg_files = glob.glob(os.path.join(textgrid_dir, '*.TextGrid'))
if not tg_files:
self._log("未找到 TextGrid 文件")
return [], set()
self._log(f"处理 {len(tg_files)} 个 TextGrid 文件")
oto_entries = []
wav_files = set()
for tg_path in tg_files:
basename = os.path.basename(tg_path).replace('.TextGrid', '')
wav_name = basename + '.wav'
wav_path = os.path.join(slices_dir, wav_name)
if not os.path.exists(wav_path):
continue
try:
info = sf.info(wav_path)
wav_duration_ms = info.duration * 1000
except Exception:
continue
wav_files.add(wav_name)
try:
tg = textgrid.TextGrid.fromFile(tg_path)
except Exception:
continue
# 查找 words 层和 phones 层
words_tier = None
phones_tier = None
for tier in tg:
name_lower = tier.name.lower()
if name_lower in ('words', 'word'):
words_tier = tier
elif name_lower in ('phones', 'phone'):
phones_tier = tier
# 如果没找到,按顺序取
if words_tier is None and len(tg) >= 1:
words_tier = tg[0]
if phones_tier is None and len(tg) >= 2:
phones_tier = tg[1]
if phones_tier is None:
continue
# 提取 CV 对,使用 words 层限制配对范围
entries = self._extract_cv_pairs(
words_tier, phones_tier, wav_name, wav_duration_ms,
language, use_hiragana, overlap_ratio
)
oto_entries.extend(entries)
# 如果启用 CVVC 模式,额外提取 VC 对
if cvvc_mode:
vc_entries = self._extract_vc_pairs(
words_tier, phones_tier, wav_name, wav_duration_ms,
language, use_hiragana,
vc_offset_ratio, vc_overlap_ratio, vc_separator
)
oto_entries.extend(vc_entries)
return oto_entries, wav_files
def _extract_cv_pairs(
self,
words_tier,
phones_tier,
wav_name: str,
wav_duration_ms: float,
language: str,
use_hiragana: bool,
overlap_ratio: float
) -> List[Dict]:
"""
从 phones 层提取音节(可能包含辅音+元音+韵尾)
使用 words 层限制配对范围,确保音素属于同一个字
"""
entries = []
# 构建 word 时间范围列表
word_ranges = []
if words_tier:
for interval in words_tier:
text = interval.mark.strip()
if text and text not in SKIP_MARKS:
word_ranges.append((interval.minTime, interval.maxTime))
def get_word_range(time: float) -> Optional[Tuple[float, float]]:
"""获取某时间点所属的 word 范围"""
for start, end in word_ranges:
if start <= time < end:
return (start, end)
return None
def same_word(time1: float, time2: float) -> bool:
"""判断两个时间点是否在同一个 word 内"""
if not word_ranges:
return True # 没有 words 层时不限制
range1 = get_word_range(time1)
range2 = get_word_range(time2)
return range1 is not None and range1 == range2
intervals = list(phones_tier)
i = 0
while i < len(intervals):
interval = intervals[i]
phone = interval.mark.strip()
if phone in SKIP_MARKS:
i += 1
continue
start_ms = interval.minTime * 1000
end_ms = interval.maxTime * 1000
# 中文音节结构:(辅音) + (介音) + 元音 + (韵尾)
if language in ('chinese', 'zh', 'mandarin'):
syllable_phones = []
syllable_start = start_ms
syllable_end = end_ms
consonant_duration = 0
# 1. 检查是否有声母(辅音)
if is_consonant(phone, language):
syllable_phones.append(phone)
consonant_duration = end_ms - start_ms
i += 1
# 检查下一个音素
if i < len(intervals):
next_interval = intervals[i]
next_phone = next_interval.mark.strip()
if next_phone not in SKIP_MARKS and same_word(interval.minTime, next_interval.minTime):
phone = next_phone
end_ms = next_interval.maxTime * 1000
syllable_end = end_ms
else:
# 只有辅音,没有元音,跳过
continue
else:
# 只有辅音,没有元音,跳过
continue
# 2. 检查是否有介音(j, w, ɥ)
phone_base = _strip_tone(phone)
if phone_base in CHINESE_MEDIALS:
syllable_phones.append(phone)
i += 1
# 检查下一个音素(必须是元音)
if i < len(intervals):
next_interval = intervals[i]
next_phone = next_interval.mark.strip()
if next_phone not in SKIP_MARKS and same_word(interval.minTime, next_interval.minTime):
phone = next_phone
end_ms = next_interval.maxTime * 1000
syllable_end = end_ms
else:
# 只有介音,没有元音,跳过
continue
else:
# 只有介音,没有元音,跳过
continue
# 3. 必须有韵母(元音)
if is_vowel(phone, language):
syllable_phones.append(phone)
if not consonant_duration:
# 零声母,辅音时长设为元音前30ms
consonant_duration = min(30, (end_ms - start_ms) * 0.2)
syllable_end = end_ms
i += 1
# 4. 检查是否有韵尾(n, ng, i, u)
if i < len(intervals):
next_interval = intervals[i]
next_phone = next_interval.mark.strip()
if (next_phone not in SKIP_MARKS and
same_word(interval.minTime, next_interval.minTime)):
# 检查是否是韵尾
next_phone_base = _strip_tone(next_phone)
if next_phone_base in CHINESE_CODAS:
syllable_phones.append(next_phone)
syllable_end = next_interval.maxTime * 1000
i += 1
# 5. 将音节转换为拼音
alias = self._syllable_to_pinyin(syllable_phones, language, use_hiragana)
if alias:
entry = self._calculate_oto_params(
wav_name=wav_name,
alias=alias,
offset=syllable_start,
consonant_duration=consonant_duration,
segment_end=syllable_end,
wav_duration_ms=wav_duration_ms,
overlap_ratio=overlap_ratio
)
entries.append(entry)
else:
# 不是元音,跳过
i += 1
else:
# 日语:简单的 CV 结构
if is_consonant(phone, language):
consonant = phone
consonant_start = start_ms
consonant_end = end_ms
consonant_time = interval.minTime
vowel = None
vowel_end = end_ms
if i + 1 < len(intervals):
next_interval = intervals[i + 1]
next_phone = next_interval.mark.strip()
next_time = next_interval.minTime
if (next_phone not in SKIP_MARKS and
is_vowel(next_phone, language) and
same_word(consonant_time, next_time)):
vowel = next_phone
vowel_end = next_interval.maxTime * 1000
i += 1
alias = ipa_to_alias(consonant, vowel, language, use_hiragana)
if alias:
consonant_duration = consonant_end - consonant_start
entry = self._calculate_oto_params(
wav_name=wav_name,
alias=alias,
offset=consonant_start,
consonant_duration=consonant_duration,
segment_end=vowel_end,
wav_duration_ms=wav_duration_ms,
overlap_ratio=overlap_ratio
)
entries.append(entry)
elif is_vowel(phone, language):
alias = ipa_to_alias(None, phone, language, use_hiragana)
if alias:
entry = self._calculate_oto_params(
wav_name=wav_name,
alias=alias,
offset=start_ms,
consonant_duration=min(30, (end_ms - start_ms) * 0.2),
segment_end=end_ms,
wav_duration_ms=wav_duration_ms,
overlap_ratio=overlap_ratio
)
entries.append(entry)
i += 1
return entries
def _syllable_to_pinyin(
self,
phones: List[str],
language: str,
use_hiragana: bool
) -> Optional[str]:
"""
将音素列表转换为标准汉语拼音(通用方法)
采用新的通用转换算法,支持所有标准汉语拼音音节
参数:
phones: 音素列表(带声调的 IPA 符号)
language: 语言
use_hiragana: 是否使用平假名(中文忽略此参数)
返回:
拼音字符串
"""
if not phones:
return None
# 去除声调
phones_base = [_strip_tone(p) for p in phones]
# 解析音节结构:(辅音) + (介音) + 元音 + (韵尾)
idx = 0
c = '' # 声母
m = '' # 介音
v = '' # 元音
cd = '' # 韵尾
# 1. 声母
if idx < len(phones_base) and is_consonant(phones_base[idx], language):
c = phones_base[idx]
idx += 1
# 2. 介音
if idx < len(phones_base) and phones_base[idx] in CHINESE_MEDIALS:
m = phones_base[idx]
idx += 1
# 3. 元音(必须)
if idx < len(phones_base) and is_vowel(phones_base[idx], language):
v = phones_base[idx]
idx += 1
else:
# 没有元音,无法形成音节
return None
# 4. 韵尾
if idx < len(phones_base) and phones_base[idx] in CHINESE_CODAS:
cd = phones_base[idx]
idx += 1
# 转换为拼音
c_py = CHINESE_CONSONANT_TO_PINYIN.get(c, '')
v_py = CHINESE_VOWEL_TO_PINYIN.get(v, v)
# 组合韵母
final = ''
if m == 'j':
# i 行韵母
if cd == 'n':
if v_py == 'a':
final = 'ian'
elif v_py == 'e':
final = 'in' # j + e + n = in (如 xin, yin)
else:
final = 'i' + v_py + 'n'
elif cd == 'ŋ':
if v_py == 'a':
final = 'iang'
elif v_py == 'o':
final = 'iong'
else:
final = 'i' + v_py + 'ng'
elif cd:
final = 'i' + v_py + cd
else:
if v_py == 'a':
final = 'ia'
elif v_py == 'e':
final = 'ie'
elif v_py == 'ao':
final = 'iao'
elif v_py == 'ou':
final = 'iu'
else:
final = 'i' + v_py
elif m == 'w':
# u 行韵母
if cd == 'n':
if v_py == 'a':
final = 'uan'
elif v_py == 'e':
final = 'un' # w + ə + n = un (如 shun)
else:
final = 'u' + v_py + 'n'
elif cd == 'ŋ':
if v_py == 'a':
final = 'uang'
elif v_py == 'e':
final = 'ueng'
else:
final = 'u' + v_py + 'ng'
elif cd:
final = 'u' + v_py + cd
else:
if v_py == 'a':
final = 'ua'
elif v_py == 'o':
final = 'uo'
elif v_py == 'ei':
final = 'ui' # w + ej = ui (如 shui)
elif v_py == 'ai':
final = 'uai'
else:
final = 'u' + v_py
elif m == 'ɥ':
# ü 行韵母
if cd == 'n':
if v_py == 'a':
final = 'van'
elif v_py == 'e':
final = 'vn'
else:
final = 'v' + v_py + 'n'
elif cd:
final = 'v' + v_py + cd
else:
if v_py == 'e':
final = 've'
else:
final = 'v' + v_py
else:
# 无介音
if cd == 'n':
final = v_py + 'n'
elif cd == 'ŋ':
final = v_py + 'ng'
elif cd:
final = v_py + cd
else:
final = v_py
# 组合声母和韵母
if not c_py:
# 零声母,需要添加 y/w/yu
if final.startswith('i'):
if final == 'i':
return 'yi'
elif final in ('in', 'ing'):
return 'y' + final
else:
return 'y' + final[1:]
elif final.startswith('u'):
if final == 'u':
return 'wu'
elif final == 'un':
return 'wen'
elif final in ('ueng', 'ong'):
return 'weng'
else:
return 'w' + final[1:]
elif final.startswith('v'):
if final == 'v':
return 'yu'
else:
return 'yu' + final[1:]
else:
return final
# 有声母
if c_py in ('j', 'q', 'x'):
# j/q/x + ü 系列,ü 写作 u
if final.startswith('v'):
return c_py + 'u' + final[1:]
else:
return c_py + final
elif c_py in ('n', 'l'):
# n/l + ü 系列,保持 v
return c_py + final
else:
# 其他声母 + ü,ü 写作 u
if final.startswith('v'):
return c_py + 'u' + final[1:]
else:
return c_py + final
def _extract_vc_pairs(
self,
words_tier,
phones_tier,
wav_name: str,
wav_duration_ms: float,
language: str,
use_hiragana: bool,
vc_offset_ratio: float,
vc_overlap_ratio: float,
vc_separator: str
) -> List[Dict]:
"""
从 phones 层提取元音+辅音对(VC 部)
VC 部是当前音节的韵母(V) + 下一个音节的声母(C)
用于连接两个相邻音节的过渡部分
使用 presamp.ini 中的映射规则来确定韵母和声母的对应关系
注意:VC 部的别名始终使用拼音格式,不受 use_hiragana 参数影响
参数:
words_tier: words 层
phones_tier: phones 层
wav_name: 音频文件名
wav_duration_ms: 音频总时长
language: 语言
use_hiragana: 是否使用平假名(VC 部忽略此参数,始终用拼音)
vc_offset_ratio: VC 偏移比例
vc_overlap_ratio: VC overlap 比例
vc_separator: VC 别名分隔符
返回:
VC 条目列表
"""
entries = []
if language not in ('chinese', 'zh', 'mandarin'):
# 非中文暂不支持 CVVC
return entries
# 加载 presamp.ini 映射
vowel_map, consonant_map = self._load_presamp_mapping()
if not vowel_map or not consonant_map:
self._log("警告: 无法加载 presamp.ini 映射,跳过 VC 部生成")
return entries
intervals = list(phones_tier)
# 解析所有音节,提取韵母和声母信息
syllables = []
i = 0
while i < len(intervals):
interval = intervals[i]
phone = interval.mark.strip()
if phone in SKIP_MARKS:
i += 1
continue
# 解析一个完整音节:(辅音) + (介音) + 元音 + (韵尾)
syllable_phones = []
syllable_start = interval.minTime * 1000
syllable_end = interval.maxTime * 1000
consonant_duration = 0
vowel_start = syllable_start
vowel_end = syllable_end
has_consonant = False
# 1. 检查是否有声母(辅音)
if is_consonant(phone, language):
syllable_phones.append(phone)
consonant_duration = interval.maxTime * 1000 - syllable_start
has_consonant = True
i += 1
# 检查下一个音素
if i < len(intervals):
next_interval = intervals[i]
next_phone = next_interval.mark.strip()
if next_phone not in SKIP_MARKS:
phone = next_phone
syllable_end = next_interval.maxTime * 1000
vowel_start = next_interval.minTime * 1000
else:
# 只有辅音,没有元音,跳过
continue
else:
# 只有辅音,没有元音,跳过
continue
# 2. 检查是否有介音(j, w, ɥ)
phone_base = _strip_tone(phone)
if phone_base in CHINESE_MEDIALS:
syllable_phones.append(phone)
i += 1
# 检查下一个音素(必须是元音)
if i < len(intervals):
next_interval = intervals[i]
next_phone = next_interval.mark.strip()
if next_phone not in SKIP_MARKS:
phone = next_phone
syllable_end = next_interval.maxTime * 1000
else:
# 只有介音,没有元音,跳过
continue
else:
# 只有介音,没有元音,跳过
continue
# 3. 必须有韵母(元音)
if is_vowel(phone, language):
syllable_phones.append(phone)
vowel_end = interval.maxTime * 1000
if not consonant_duration:
# 零声母,辅音时长设为元音前30ms
consonant_duration = min(30, (vowel_end - vowel_start) * 0.2)
syllable_end = vowel_end
i += 1
# 4. 检查是否有韵尾(n, ng, i, u)
if i < len(intervals):
next_interval = intervals[i]
next_phone = next_interval.mark.strip()
if next_phone not in SKIP_MARKS:
# 检查是否是韵尾
next_phone_base = _strip_tone(next_phone)
if next_phone_base in CHINESE_CODAS:
syllable_phones.append(next_phone)
syllable_end = next_interval.maxTime * 1000
vowel_end = next_interval.maxTime * 1000
i += 1
# 5. 将音节转换为拼音并保存
pinyin = self._syllable_to_pinyin(syllable_phones, language, False)
if pinyin:
# 使用 presamp.ini 映射查找韵母和声母
vowel_part = self._find_vowel_in_mapping(pinyin, vowel_map)
consonant_part = self._find_consonant_in_mapping(pinyin, consonant_map) if has_consonant else None
if vowel_part:
syllables.append({
'pinyin': pinyin,
'vowel_part': vowel_part,
'consonant_part': consonant_part,
'vowel_start': vowel_start,
'vowel_end': vowel_end,
'syllable_end': syllable_end
})
else:
# 不是元音,跳过
i += 1
# 生成 VC 对:当前音节的韵母 + 下一个音节的声母
for idx in range(len(syllables) - 1):
current = syllables[idx]
next_syl = syllables[idx + 1]
# 获取下一个音节的声母
next_consonant = next_syl.get('consonant_part')
# 如果下一个音节没有声母(零声母),跳过
if not next_consonant:
continue
# 生成 VC 别名
vc_alias = f"{current['vowel_part']}{vc_separator}{next_consonant}"
# 计算 VC 参数
entry = self._calculate_vc_params(
wav_name=wav_name,
alias=vc_alias,
vowel_start_ms=current['vowel_start'],
vowel_end_ms=current['vowel_end'],
consonant_end_ms=next_syl['syllable_end'],
wav_duration_ms=wav_duration_ms,
vc_offset_ratio=vc_offset_ratio,
vc_overlap_ratio=vc_overlap_ratio
)
entries.append(entry)
return entries
def _load_presamp_mapping(self) -> Tuple[Dict[str, str], Dict[str, str]]:
"""
加载中文 CVVC 韵母和声母映射(内置数据)
返回:
(韵母映射字典, 声母映射字典)
韵母映射: {完整拼音: 韵母标识}
声母映射: {完整拼音: 声母标识}
"""
vowel_map = {} # {拼音: 韵母标识}
consonant_map = {} # {拼音: 声母标识}
# 内置韵母映射数据(来自 presamp.ini [VOWEL] 部分)
vowel_data = {
'a': ['a', 'ba', 'pa', 'ma', 'fa', 'da', 'ta', 'na', 'la', 'ga', 'ka', 'ha', 'zha', 'cha', 'sha', 'za', 'ca', 'sa', 'ya', 'lia', 'jia', 'qia', 'xia', 'wa', 'gua', 'kua', 'hua', 'zhua', 'shua', 'dia'],
'ai': ['ai', 'bai', 'pai', 'mai', 'dai', 'tai', 'nai', 'lai', 'gai', 'kai', 'hai', 'zhai', 'chai', 'shai', 'zai', 'cai', 'sai', 'wai', 'guai', 'kuai', 'huai', 'zhuai', 'chuai', 'shuai'],
'an': ['an', 'ban', 'pan', 'man', 'fan', 'dan', 'tan', 'nan', 'lan', 'gan', 'kan', 'han', 'zhan', 'chan', 'shan', 'ran', 'zan', 'can', 'san', 'wan', 'duan', 'tuan', 'nuan', 'luan', 'guan', 'kuan', 'huan', 'zhuan', 'chuan', 'shuan', 'ruan', 'zuan', 'cuan', 'suan'],
'ang': ['ang', 'bang', 'pang', 'mang', 'fang', 'dang', 'tang', 'nang', 'lang', 'gang', 'kang', 'hang', 'zhang', 'chang', 'shang', 'rang', 'zang', 'cang', 'sang', 'yang', 'liang', 'jiang', 'qiang', 'xiang', 'wang', 'guang', 'kuang', 'huang', 'zhuang', 'chuang', 'shuang', 'niang'],
'ao': ['ao', 'bao', 'pao', 'mao', 'dao', 'tao', 'nao', 'lao', 'gao', 'kao', 'hao', 'zhao', 'chao', 'shao', 'rao', 'zao', 'cao', 'sao', 'yao', 'biao', 'piao', 'miao', 'diao', 'tiao', 'niao', 'liao', 'jiao', 'qiao', 'xiao'],
'e': ['e', 'me', 'de', 'te', 'ne', 'le', 'ge', 'ke', 'he', 'zhe', 'che', 'she', 're', 'ze', 'ce', 'se'],
'e0': ['ye', 'bie', 'pie', 'mie', 'die', 'tie', 'nie', 'lie', 'jie', 'qie', 'xie', 'yue', 'nue', 'lue', 'jue', 'que', 'xue'],
'ei': ['ei', 'bei', 'pei', 'mei', 'fei', 'dei', 'tei', 'nei', 'lei', 'gei', 'kei', 'hei', 'zhei', 'shei', 'zei', 'wei', 'dui', 'tui', 'gui', 'kui', 'hui', 'zhui', 'chui', 'shui', 'rui', 'zui', 'cui', 'sui'],
'en': ['en', 'ben', 'pen', 'men', 'fen', 'nen', 'gen', 'ken', 'hen', 'zhen', 'chen', 'shen', 'ren', 'zen', 'cen', 'sen', 'wen', 'dun', 'tun', 'lun', 'gun', 'kun', 'hun', 'zhun', 'chun', 'shun', 'run', 'zun', 'cun', 'sun'],
'en0': ['yan', 'bian', 'pian', 'mian', 'dian', 'tian', 'nian', 'lian', 'jian', 'qian', 'xian', 'yuan', 'juan', 'quan', 'xuan'],
'eng': ['beng', 'peng', 'meng', 'feng', 'deng', 'teng', 'neng', 'leng', 'geng', 'keng', 'heng', 'weng', 'zheng', 'cheng', 'sheng', 'reng', 'zeng', 'ceng', 'seng'],
'er': ['er'],
'i': ['bi', 'pi', 'mi', 'di', 'ti', 'ni', 'li', 'ji', 'qi', 'xi', 'yi'],
'in': ['yin', 'bin', 'pin', 'min', 'nin', 'lin', 'jin', 'qin', 'xin'],
'ing': ['ying', 'bing', 'ping', 'ming', 'ding', 'ting', 'ning', 'ling', 'jing', 'qing', 'xing'],
'i0': ['zi', 'ci', 'si'],
'ir': ['zhi', 'chi', 'shi', 'ri'],
'o': ['bo', 'po', 'mo', 'fo', 'wo', 'duo', 'tuo', 'nuo', 'luo', 'guo', 'kuo', 'huo', 'zhuo', 'chuo', 'shuo', 'ruo', 'zuo', 'cuo', 'suo'],
'ong': ['dong', 'tong', 'nong', 'long', 'gong', 'kong', 'hong', 'zhong', 'chong', 'rong', 'zong', 'cong', 'song', 'yong', 'jiong', 'qiong', 'xiong'],
'ou': ['ou', 'pou', 'mou', 'fou', 'dou', 'tou', 'lou', 'gou', 'kou', 'hou', 'zhou', 'chou', 'shou', 'rou', 'zou', 'cou', 'sou', 'you', 'miu', 'diu', 'niu', 'liu', 'jiu', 'qiu', 'xiu'],
'u': ['bu', 'pu', 'mu', 'fu', 'du', 'tu', 'nu', 'lu', 'gu', 'ku', 'hu', 'zhu', 'chu', 'shu', 'ru', 'zu', 'cu', 'su', 'wu'],
'v': ['yu', 'nv', 'lv', 'ju', 'qu', 'xu'],
'vn': ['yun', 'jun', 'qun', 'xun'],
}
# 内置声母映射数据(来自 presamp.ini [CONSONANT] 部分)
consonant_data = {
'b': ['ba', 'bai', 'ban', 'bang', 'bao', 'biao', 'bie', 'bei', 'ben', 'bian', 'beng', 'bi', 'bin', 'bing', 'bo', 'bu'],
'p': ['pa', 'pai', 'pan', 'pang', 'pao', 'piao', 'pie', 'pei', 'pen', 'pian', 'peng', 'pi', 'pin', 'ping', 'po', 'pou', 'pu'],
'm': ['ma', 'mai', 'man', 'mang', 'mao', 'me', 'mei', 'men', 'meng', 'mo', 'mou', 'mu'],
'f': ['fa', 'fan', 'fang', 'fei', 'fen', 'feng', 'fo', 'fou', 'fu'],
'd': ['da', 'dia', 'dai', 'dan', 'duan', 'dang', 'dao', 'diao', 'de', 'die', 'dei', 'dui', 'dun', 'dian', 'deng', 'di', 'ding', 'duo', 'dong', 'dou', 'diu', 'du'],
't': ['ta', 'tai', 'tan', 'tuan', 'tang', 'tao', 'tiao', 'te', 'tie', 'tei', 'tui', 'tun', 'tian', 'teng', 'ti', 'ting', 'tuo', 'tong', 'tou', 'tu'],
'n': ['na', 'nai', 'nan', 'nuan', 'nang', 'nao', 'ne', 'nue', 'nei', 'nen', 'neng', 'nuo', 'nong', 'nu', 'nv'],
'l': ['la', 'lai', 'lan', 'luan', 'lang', 'lao', 'le', 'lue', 'lei', 'lun', 'leng', 'luo', 'long', 'lou', 'lu', 'lv'],
'g': ['ga', 'gua', 'gai', 'guai', 'gan', 'guan', 'gang', 'guang', 'gao', 'ge', 'gei', 'gui', 'gen', 'gun', 'geng', 'guo', 'gong', 'gou', 'gu'],
'k': ['ka', 'kua', 'kai', 'kuai', 'kan', 'kuan', 'kang', 'kuang', 'kao', 'ke', 'kei', 'kui', 'ken', 'kun', 'keng', 'kuo', 'kong', 'kou', 'ku'],
'h': ['ha', 'hai', 'han', 'hang', 'hao', 'he', 'hei', 'hen', 'heng', 'hong', 'hou'],
'zh': ['zha', 'zhua', 'zhai', 'zhuai', 'zhan', 'zhuan', 'zhang', 'zhuang', 'zhao', 'zhe', 'zhei', 'zhui', 'zhen', 'zhun', 'zheng', 'zhi', 'zhuo', 'zhong', 'zhou', 'zhu'],
'ch': ['cha', 'chai', 'chuai', 'chan', 'chuan', 'chang', 'chuang', 'chao', 'che', 'chui', 'chen', 'chun', 'cheng', 'chi', 'chuo', 'chong', 'chou', 'chu'],
'sh': ['sha', 'shai', 'shan', 'shang', 'shao', 'she', 'shei', 'shen', 'sheng', 'shi', 'shou'],
'z': ['za', 'zai', 'zan', 'zuan', 'zang', 'zao', 'ze', 'zei', 'zui', 'zen', 'zun', 'zeng', 'zi', 'zuo', 'zong', 'zou', 'zu'],
'c': ['ca', 'cai', 'can', 'cuan', 'cang', 'cao', 'ce', 'cui', 'cen', 'cun', 'ceng', 'ci', 'cuo', 'cong', 'cou', 'cu'],
's': ['sa', 'sai', 'san', 'sang', 'sao', 'se', 'sen', 'seng', 'si', 'song', 'sou'],
'y': ['ya', 'yang', 'yao', 'ye', 'yan', 'yi', 'yin', 'ying', 'yong', 'you'],
'ly': ['lia', 'liang', 'liao', 'lie', 'lian', 'li', 'lin', 'ling', 'liu'],
'j': ['jia', 'jiang', 'jiao', 'jie', 'jue', 'jian', 'juan', 'ji', 'jin', 'jing', 'jiong', 'jiu', 'ju', 'jun'],
'q': ['qia', 'qiang', 'qiao', 'qie', 'que', 'qian', 'quan', 'qi', 'qin', 'qing', 'qiong', 'qiu', 'qu', 'qun'],
'xy': ['xia', 'xiang', 'xiao', 'xie', 'xian', 'xi', 'xin', 'xing', 'xiong', 'xiu'],
'w': ['wa', 'wai', 'wan', 'wang', 'wei', 'wen', 'weng', 'wo', 'wu'],
'hw': ['hua', 'huai', 'huan', 'huang', 'hui', 'hun', 'huo', 'hu'],
'shw': ['shua', 'shuai', 'shuan', 'shuang', 'shui', 'shun', 'shuo', 'shu'],
'r': ['ran', 'ruan', 'rang', 'rao', 're', 'rui', 'ren', 'run', 'reng', 'ri', 'ruo', 'rong', 'rou', 'ru'],
'sw': ['suan', 'sui', 'sun', 'suo', 'su'],
'ny': ['niang', 'niao', 'nie', 'nian', 'ni', 'nin', 'ning', 'niu'],
'my': ['miao', 'mie', 'mian', 'mi', 'min', 'ming', 'miu'],
'v': ['yu', 'yue', 'yuan', 'yun'],
'xw': ['xue', 'xuan', 'xu', 'xun'],
}
# 构建韵母映射
for vowel_id, pinyins in vowel_data.items():
for pinyin in pinyins:
vowel_map[pinyin] = vowel_id
# 构建声母映射
for consonant_id, pinyins in consonant_data.items():
for pinyin in pinyins:
consonant_map[pinyin] = consonant_id
self._log(f"加载内置 CVVC 映射: {len(vowel_map)} 个韵母映射, {len(consonant_map)} 个声母映射")
return vowel_map, consonant_map
def _find_vowel_in_mapping(self, pinyin: str, vowel_map: Dict[str, str]) -> Optional[str]:
"""
在韵母映射中查找拼音对应的韵母标识
参数:
pinyin: 完整拼音
vowel_map: 韵母映射字典
返回:
韵母标识,如果未找到则返回 None
"""
return vowel_map.get(pinyin)
def _find_consonant_in_mapping(self, pinyin: str, consonant_map: Dict[str, str]) -> Optional[str]:
"""
在声母映射中查找拼音对应的声母标识
参数:
pinyin: 完整拼音
consonant_map: 声母映射字典
返回:
声母标识,如果未找到则返回 None
"""
return consonant_map.get(pinyin)
def _calculate_oto_params(
self,
wav_name: str,
alias: str,
offset: float,
consonant_duration: float,
segment_end: float,
wav_duration_ms: float,
overlap_ratio: float
) -> Dict:
"""
计算 oto.ini 参数
oto.ini 格式: wav=alias,offset,consonant,cutoff,preutterance,overlap
- offset: 从音频开头跳过的毫秒数
- consonant: 不被拉伸的区域长度
- cutoff: 负值,表示这个音素的总时长(从 offset 开始)
- preutterance: 先行发声
- overlap: 与前一音符的交叉淡化区域
"""
segment_duration = segment_end - offset
preutterance = consonant_duration
overlap = preutterance * overlap_ratio
# cutoff 为负值,表示音素的总时长
cutoff = -segment_duration
return {
"wav_name": wav_name,
"alias": alias,
"offset": round(offset, 1),
"consonant": round(consonant_duration, 1),
"cutoff": round(cutoff, 1),
"preutterance": round(preutterance, 1),
"overlap": round(overlap, 1),
"segment_duration": segment_duration, # 用于排序
}
def _calculate_vc_params(
self,
wav_name: str,
alias: str,
vowel_start_ms: float,
vowel_end_ms: float,
consonant_end_ms: float,
wav_duration_ms: float,
vc_offset_ratio: float,
vc_overlap_ratio: float
) -> Dict:
"""
计算 VC 部的 oto.ini 参数
VC 部从元音后半段开始,到辅音结束
参数:
wav_name: 音频文件名
alias: VC 别名
vowel_start_ms: 元音开始时间
vowel_end_ms: 元音结束时间(即辅音开始时间)
consonant_end_ms: 辅音结束时间
wav_duration_ms: 音频总时长
vc_offset_ratio: VC 偏移比例
vc_overlap_ratio: VC overlap 比例
返回:
oto 参数字典
"""
vowel_duration = vowel_end_ms - vowel_start_ms
# offset: 元音后半段位置
offset = vowel_end_ms - vowel_duration * vc_offset_ratio
# 总时长(从 offset 到辅音结束)
segment_duration = consonant_end_ms - offset
# preutterance: 从 offset 到辅音开始(即元音结束)的距离
preutterance = vowel_end_ms - offset
# consonant: 固定区域,较短
consonant = min(30, segment_duration * 0.3)
# overlap: 较大,平滑过渡
overlap = preutterance * vc_overlap_ratio
# cutoff: 负值,表示总时长
cutoff = -segment_duration
return {
"wav_name": wav_name,
"alias": alias,
"offset": round(offset, 1),
"consonant": round(consonant, 1),
"cutoff": round(cutoff, 1),
"preutterance": round(preutterance, 1),
"overlap": round(overlap, 1),
"segment_duration": segment_duration,
"is_vc": True # 标记为 VC 部
}
def _filter_by_alias(
self,
entries: List[Dict],
max_samples: int,
naming_rule: str,
first_naming_rule: str,
slices_dir: str,
enabled_metrics: List[str]
) -> Tuple[List[Dict], set]:
"""按别名分组,使用质量评分筛选最佳样本,并添加编号"""
# 过滤空别名
valid_entries = [e for e in entries if e.get("alias") and e["alias"].strip()]
# 按基础别名分组
alias_groups: Dict[str, List[Dict]] = defaultdict(list)
for entry in valid_entries:
alias_groups[entry["alias"]].append(entry)
# 判断是否需要加载音频计算质量分数
need_audio_scoring = any(m in enabled_metrics for m in ["rms", "f0"])
filtered = []
used_wavs = set()
for base_alias, group in alias_groups.items():
# 计算质量分数
if need_audio_scoring:
scored_group = self._score_entries(group, slices_dir, enabled_metrics)
else:
# 仅使用时长评分
from ..quality_scorer import duration_score
for entry in group:
duration = entry["segment_duration"] / 1000 # 转换为秒
entry["quality_score"] = duration_score(duration)
scored_group = group
# 按质量分数排序(降序)
sorted_group = sorted(scored_group, key=lambda x: -x.get("quality_score", 0))
# 保留前 N 个,并应用命名规则
for idx, entry in enumerate(sorted_group[:max_samples]):
# 使用基类方法应用命名规则
if idx == 0 and first_naming_rule:
final_alias = self.apply_naming_rule(first_naming_rule, base_alias, idx)
else:
final_alias = self.apply_naming_rule(naming_rule, base_alias, idx)
entry["alias"] = final_alias
filtered.append(entry)
used_wavs.add(entry["wav_name"])
return filtered, used_wavs
def _score_entries(
self,
entries: List[Dict],
slices_dir: str,
enabled_metrics: List[str]
) -> List[Dict]:
"""为条目计算质量分数"""
import soundfile as sf
from ..quality_scorer import QualityScorer
scorer = QualityScorer(enabled_metrics=enabled_metrics)
# 缓存已加载的音频
audio_cache: Dict[str, Tuple] = {}
for entry in entries:
wav_name = entry["wav_name"]
wav_path = os.path.join(slices_dir, wav_name)
try:
# 加载或使用缓存的音频
if wav_name not in audio_cache:
audio, sr = sf.read(wav_path)
if len(audio.shape) > 1:
audio = audio.mean(axis=1)
audio_cache[wav_name] = (audio, sr)
else:
audio, sr = audio_cache[wav_name]
# 提取片段(根据 offset 和 segment_duration)
offset_samples = int(entry["offset"] / 1000 * sr)
duration_samples = int(entry["segment_duration"] / 1000 * sr)
segment = audio[offset_samples:offset_samples + duration_samples]
if len(segment) > 0:
scores = scorer.score(segment, sr)
entry["quality_score"] = scores.get("combined", 0.5)
else:
entry["quality_score"] = 0.5
except Exception as e:
logger.warning(f"评分失败 {wav_name}: {e}")
entry["quality_score"] = 0.5
return entries
def _copy_wav_files(
self,
wav_files: set,
slices_dir: str,
export_dir: str,
encoding: str = "shift_jis"
) -> Tuple[int, Dict[str, str]]:
"""
复制音频文件到导出目录
参数:
wav_files: 需要复制的文件名集合
slices_dir: 源目录
export_dir: 目标目录
encoding: 目标编码,用于检测文件名是否合法
返回:
(复制数量, 文件名映射表 {原文件名: 新文件名})
"""
copied = 0
filename_map: Dict[str, str] = {}
used_names: set = set()
sanitized_count = 0
for wav_name in wav_files:
src = os.path.join(slices_dir, wav_name)
if not os.path.exists(src):
continue
# 检测文件名是否能用指定编码表示
if self._is_filename_valid(wav_name, encoding):
new_name = wav_name
else:
new_name = self._sanitize_filename(wav_name, used_names)
sanitized_count += 1
used_names.add(new_name)
filename_map[wav_name] = new_name
dst = os.path.join(export_dir, new_name)
shutil.copyfile(src, dst)
copied += 1
if sanitized_count > 0:
self._log(f"已将 {sanitized_count} 个文件名转换为拼音(原文件名无法用 {encoding} 编码)")
return copied, filename_map
def _is_filename_valid(self, filename: str, encoding: str) -> bool:
"""
检测文件名是否合法(能否用指定编码表示)
参数:
filename: 文件名
encoding: 目标编码
返回:
True 表示文件名合法,False 表示需要转换
"""
try:
filename.encode(encoding)
return True
except UnicodeEncodeError:
return False
def _sanitize_filename(self, filename: str, used_names: set) -> str:
"""
清理文件名:中文转拼音 + 特殊字符清理 + 防冲突
参数:
filename: 原文件名
used_names: 已使用的文件名集合(用于防冲突)
返回:
清理后的文件名
"""
from pypinyin import lazy_pinyin
import re
# 分离文件名和扩展名
name, ext = os.path.splitext(filename)
# 中文转拼音
pinyin_parts = lazy_pinyin(name)
sanitized = ''.join(pinyin_parts)
# 清理特殊字符,只保留字母、数字、下划线、连字符
sanitized = re.sub(r'[^a-zA-Z0-9_\-]', '_', sanitized)
# 合并连续下划线
sanitized = re.sub(r'_+', '_', sanitized)
# 去除首尾下划线
sanitized = sanitized.strip('_')
# 如果为空,使用默认名
if not sanitized:
sanitized = 'audio'
# 防冲突:添加数字后缀
base_name = sanitized
counter = 1
while f"{sanitized}{ext}" in used_names:
sanitized = f"{base_name}_{counter}"
counter += 1
return f"{sanitized}{ext}"
def _write_oto_ini(
self,
entries: List[Dict],
output_path: str,
encoding: str,
filename_map: Optional[Dict[str, str]] = None
):
"""
写入 oto.ini 文件
参数:
entries: oto 条目列表
output_path: 输出路径
encoding: 文件编码
filename_map: 文件名映射表(原文件名 -> 新文件名)
"""
lines = []
for entry in entries:
# 跳过空别名
alias = entry.get("alias", "")
if not alias or not alias.strip():
logger.warning(f"跳过空别名: {entry.get('wav_name', 'unknown')}")
continue
# 应用文件名映射
wav_name = entry["wav_name"]
if filename_map and wav_name in filename_map:
wav_name = filename_map[wav_name]
line = "{wav}={alias},{offset},{consonant},{cutoff},{preutterance},{overlap}".format(
wav=wav_name,
alias=alias,
offset=entry["offset"],
consonant=entry["consonant"],
cutoff=entry["cutoff"],
preutterance=entry["preutterance"],
overlap=entry["overlap"]
)
lines.append(line)
# 按 wav 文件名 + 别名排序
lines.sort(key=lambda x: (x.split('=')[0], x.split('=')[1].split(',')[0]))
with open(output_path, 'w', encoding=encoding) as f:
f.write('\n'.join(lines))
def _write_character_txt(
self,
character_name: str,
output_path: str,
encoding: str
):
"""写入 character.txt 文件,用于 UTAU 识别音源名称
参数:
character_name: 角色名称(可以是用户自定义的名称或音源名称)
output_path: 输出路径
encoding: 文件编码
注意:当角色名称包含无法用指定编码表示的字符时,
自动将名称转换为拼音/罗马音。
"""
name_to_write = character_name
# 检测是否能用指定编码
try:
character_name.encode(encoding)
except UnicodeEncodeError:
# 无法编码,转换为拼音
from pypinyin import lazy_pinyin
pinyin_name = ''.join(lazy_pinyin(character_name))
logger.warning(f"角色名称 '{character_name}' 无法用 {encoding} 编码,已转换为拼音: {pinyin_name}")
self._log(f"角色名称 '{character_name}' 无法用 {encoding} 编码,已转换为拼音: {pinyin_name}")
name_to_write = pinyin_name
with open(output_path, 'w', encoding=encoding) as f:
f.write(f"name={name_to_write}")
# ==================== 自动拼字功能 ====================
def _auto_combine_phonemes(
self,
all_entries: List[Dict],
filtered_entries: List[Dict],
slices_dir: str,
export_dir: str,
language: str,
use_hiragana: bool,
overlap_ratio: float,
crossfade_ms: int,
first_naming_rule: str,
fuzzy_phoneme: bool = False
) -> Tuple[List[Dict], set]:
"""
自动拼字:用已有音素拼接生成缺失的音素组合
参数:
all_entries: 所有原始 oto 条目(用于提取音素片段)
filtered_entries: 已筛选的条目(用于确定已有别名)
slices_dir: 切片目录
export_dir: 导出目录
language: 语言
use_hiragana: 是否使用平假名
overlap_ratio: overlap 比例
crossfade_ms: 交叉淡化时长
first_naming_rule: 首个样本命名规则
fuzzy_phoneme: 是否启用模糊拼字(仅中文有效)
返回:
(新生成的条目列表, 新生成的 wav 文件名集合)
"""
import numpy as np
import soundfile as sf
# 步骤1: 收集已有别名
existing_aliases = set()
for entry in filtered_entries:
# 提取基础别名(去除序号后缀)
alias = entry.get("alias", "")
if alias:
existing_aliases.add(alias)
self._log(f"已有 {len(existing_aliases)} 个别名")
# 步骤2: 从原始条目中提取最佳辅音和元音片段
consonant_segments, vowel_segments = self._collect_phoneme_segments(
all_entries, slices_dir, language
)
self._log(f"收集到 {len(consonant_segments)} 个辅音, {len(vowel_segments)} 个元音")
if not consonant_segments or not vowel_segments:
self._log("音素不足,跳过自动拼字")
return [], set()
# 步骤3: 生成候选组合并过滤
# 模糊拼字仅对中文生效
enable_fuzzy = fuzzy_phoneme and language in ('chinese', 'zh', 'mandarin')
candidates = self._generate_candidates(
consonant_segments, vowel_segments,
existing_aliases, language, use_hiragana,
enable_fuzzy
)
if not candidates:
self._log("无缺失的有效组合")
return [], set()
self._log(f"发现 {len(candidates)} 个缺失组合,开始拼接...")
# 步骤4: 执行音频拼接
new_entries = []
new_wavs = set()
success_count = 0
fail_count = 0
for candidate in candidates:
try:
entry, wav_name = self._combine_and_save(
candidate,
slices_dir,
export_dir,
overlap_ratio,
crossfade_ms,
first_naming_rule
)
if entry:
new_entries.append(entry)
new_wavs.add(wav_name)
success_count += 1
except Exception as e:
logger.warning(f"拼接失败 {candidate['alias']}: {e}")
fail_count += 1
if fail_count > 0:
self._log(f"拼接完成: 成功 {success_count}, 失败 {fail_count}")
return new_entries, new_wavs
def _collect_phoneme_segments(
self,
entries: List[Dict],
slices_dir: str,
language: str
) -> Tuple[Dict[str, Dict], Dict[str, Dict]]:
"""
从条目中收集辅音和元音片段信息
返回:
(辅音字典, 元音字典)
每个字典: {IPA音素: {wav_path, offset_ms, duration_ms, quality_score}}
"""
import soundfile as sf
consonant_segments: Dict[str, List[Dict]] = defaultdict(list)
vowel_segments: Dict[str, List[Dict]] = defaultdict(list)
for entry in entries:
wav_name = entry.get("wav_name", "")
wav_path = os.path.join(slices_dir, wav_name)
if not os.path.exists(wav_path):
continue
# 从条目中提取原始音素信息(如果有)
# 这里需要重新解析,因为原始条目可能没有保存 IPA 信息
# 我们使用 alias 反推(简化处理)
alias = entry.get("alias", "")
offset = entry.get("offset", 0)
consonant_dur = entry.get("consonant", 0)
segment_dur = entry.get("segment_duration", 0)
quality = entry.get("quality_score", 0.5)
# 尝试分离辅音和元音部分
c_part, v_part = self._split_alias_to_cv(alias, language)
if c_part:
consonant_segments[c_part].append({
"wav_path": wav_path,
"offset_ms": offset,
"duration_ms": consonant_dur,
"quality_score": quality,
"ipa": c_part
})
if v_part:
# 元音从辅音结束位置开始
v_offset = offset + consonant_dur
v_duration = segment_dur - consonant_dur
if v_duration > 0:
vowel_segments[v_part].append({
"wav_path": wav_path,
"offset_ms": v_offset,
"duration_ms": v_duration,
"quality_score": quality,
"ipa": v_part
})
# 选择最佳音素
# 辅音:从质量前5中选择时长最接近中位数的(避免过长或过短)
# 元音:从质量前5中选择时长最长的(避免UTAU过度拉伸)
best_consonants = {}
for ipa, segments in consonant_segments.items():
if segments:
best_consonants[ipa] = self._select_best_consonant(segments)
best_vowels = {}
for ipa, segments in vowel_segments.items():
if segments:
best_vowels[ipa] = self._select_best_vowel(segments)
return best_consonants, best_vowels
def _select_best_consonant(self, segments: List[Dict]) -> Dict:
"""
选择最佳辅音片段
策略:从质量排名前5中选择时长最接近中位数的
(辅音不宜过长也不宜过短)
"""
# 按质量排序,取前5
sorted_by_quality = sorted(segments, key=lambda x: -x["quality_score"])
top_candidates = sorted_by_quality[:5]
if len(top_candidates) == 1:
return top_candidates[0]
# 计算这些候选的时长中位数
durations = [s["duration_ms"] for s in top_candidates]
durations.sort()
median_duration = durations[len(durations) // 2]
# 选择最接近中位数的
best = min(top_candidates, key=lambda x: abs(x["duration_ms"] - median_duration))
return best
def _select_best_vowel(self, segments: List[Dict]) -> Dict:
"""
选择最佳元音片段
策略:从质量排名前5中选择时长最长的
(元音过短会导致UTAU过度拉伸)
"""
# 按质量排序,取前5
sorted_by_quality = sorted(segments, key=lambda x: -x["quality_score"])
top_candidates = sorted_by_quality[:5]
# 从中选择时长最长的
best = max(top_candidates, key=lambda x: x["duration_ms"])
return best
def _split_alias_to_cv(
self,
alias: str,
language: str
) -> Tuple[Optional[str], Optional[str]]:
"""
将别名拆分为辅音和元音部分
参数:
alias: 别名(拼音、罗马音或平假名)
language: 语言
返回:
(辅音部分, 元音部分) - 始终返回罗马音格式
"""
if not alias:
return None, None
# 如果是平假名,先转换为罗马音
alias_to_split = self._hiragana_to_romaji(alias)
if alias_to_split is None:
alias_to_split = alias.lower()
if language in ('chinese', 'zh', 'mandarin'):
# 中文拼音辅音列表(按长度降序排列以优先匹配长的)
consonants = [
'zh', 'ch', 'sh', 'ng',
'b', 'p', 'm', 'f',
'd', 't', 'n', 'l',
'g', 'k', 'h',
'j', 'q', 'x',
'z', 'c', 's', 'r',
'y', 'w'
]
else:
# 日语罗马音辅音
consonants = [
'ch', 'sh', 'ts', 'ny',
'ky', 'gy', 'py', 'by', 'my', 'ry', 'hy',
'k', 'g', 's', 'z', 't', 'd', 'n', 'h', 'b', 'p', 'm', 'r', 'w', 'y', 'f', 'j'
]
# 尝试匹配辅音
for c in consonants:
if alias_to_split.startswith(c):
vowel = alias_to_split[len(c):]
if vowel: # 确保有元音部分
return c, vowel
else:
return c, None
# 没有辅音,整个是元音
return None, alias_to_split
def _hiragana_to_romaji(self, text: str) -> Optional[str]:
"""
将平假名转换为罗马音
参数:
text: 平假名文本
返回:
罗马音,如果无法转换则返回 None
"""
# 平假名到罗马音映射(ROMAJI_TO_HIRAGANA 的反向映射)
hiragana_to_romaji_map = {
# 基本元音
'あ': 'a', 'い': 'i', 'う': 'u', 'え': 'e', 'お': 'o',
# か行
'か': 'ka', 'き': 'ki', 'く': 'ku', 'け': 'ke', 'こ': 'ko',
# さ行
'さ': 'sa', 'し': 'shi', 'す': 'su', 'せ': 'se', 'そ': 'so',
# た行
'た': 'ta', 'ち': 'chi', 'つ': 'tsu', 'て': 'te', 'と': 'to',
# な行
'な': 'na', 'に': 'ni', 'ぬ': 'nu', 'ね': 'ne', 'の': 'no',
# は行
'は': 'ha', 'ひ': 'hi', 'ふ': 'fu', 'へ': 'he', 'ほ': 'ho',
# ま行
'ま': 'ma', 'み': 'mi', 'む': 'mu', 'め': 'me', 'も': 'mo',
# や行
'や': 'ya', 'ゆ': 'yu', 'よ': 'yo',
# ら行
'ら': 'ra', 'り': 'ri', 'る': 'ru', 'れ': 're', 'ろ': 'ro',
# わ行
'わ': 'wa', 'を': 'wo', 'ん': 'n',
# が行
'が': 'ga', 'ぎ': 'gi', 'ぐ': 'gu', 'げ': 'ge', 'ご': 'go',
# ざ行
'ざ': 'za', 'じ': 'ji', 'ず': 'zu', 'ぜ': 'ze', 'ぞ': 'zo',
# だ行
'だ': 'da', 'ぢ': 'di', 'づ': 'du', 'で': 'de', 'ど': 'do',
# ば行
'ば': 'ba', 'び': 'bi', 'ぶ': 'bu', 'べ': 'be', 'ぼ': 'bo',
# ぱ行
'ぱ': 'pa', 'ぴ': 'pi', 'ぷ': 'pu', 'ぺ': 'pe', 'ぽ': 'po',
# 拗音
'きゃ': 'kya', 'きゅ': 'kyu', 'きょ': 'kyo',
'しゃ': 'sha', 'しゅ': 'shu', 'しょ': 'sho',
'ちゃ': 'cha', 'ちゅ': 'chu', 'ちょ': 'cho',
'にゃ': 'nya', 'にゅ': 'nyu', 'にょ': 'nyo',
'ひゃ': 'hya', 'ひゅ': 'hyu', 'ひょ': 'hyo',
'みゃ': 'mya', 'みゅ': 'myu', 'みょ': 'myo',
'りゃ': 'rya', 'りゅ': 'ryu', 'りょ': 'ryo',
'ぎゃ': 'gya', 'ぎゅ': 'gyu', 'ぎょ': 'gyo',
'じゃ': 'ja', 'じゅ': 'ju', 'じょ': 'jo',
'びゃ': 'bya', 'びゅ': 'byu', 'びょ': 'byo',
'ぴゃ': 'pya', 'ぴゅ': 'pyu', 'ぴょ': 'pyo',
}
# 去除数字后缀
base_text = text.rstrip('0123456789')
# 直接查找
if base_text in hiragana_to_romaji_map:
return hiragana_to_romaji_map[base_text]
# 如果是纯 ASCII,直接返回小写
if base_text.isascii():
return base_text.lower()
return None
def _generate_candidates(
self,
consonants: Dict[str, Dict],
vowels: Dict[str, Dict],
existing_aliases: set,
language: str,
use_hiragana: bool,
fuzzy_phoneme: bool = False
) -> List[Dict]:
"""
生成缺失的候选组合
参数:
consonants: 可用辅音字典
vowels: 可用元音字典
existing_aliases: 已存在的别名集合
language: 语言
use_hiragana: 是否使用平假名
fuzzy_phoneme: 是否启用模糊拼字
返回:
候选列表,每个候选包含 {alias, consonant_info, vowel_info}
"""
candidates = []
# 获取有效的元音列表(用于验证组合)
if language in ('chinese', 'zh', 'mandarin'):
valid_vowels = {'a', 'o', 'e', 'i', 'u', 'v',
'ai', 'ei', 'ao', 'ou',
'an', 'en', 'ang', 'eng', 'ong',
'ia', 'ie', 'iao', 'iu', 'ian', 'in', 'iang', 'ing', 'iong',
'ua', 'uo', 'uai', 'ui', 'uan', 'un', 'uang', 'ueng',
've', 'van', 'vn', 'er'}
else:
valid_vowels = {'a', 'i', 'u', 'e', 'o'}
# 构建可用音素集合(用于模糊匹配)
available_consonants = set(consonants.keys())
available_vowels = set(vowels.keys())
# 辅音 + 元音组合
for c_alias, c_info in consonants.items():
for v_alias, v_info in vowels.items():
# 确保辅音和元音都是罗马音格式(小写ASCII)
c_romaji = c_alias.lower() if c_alias.isascii() else None
v_romaji = v_alias.lower() if v_alias.isascii() else None
# 跳过非罗马音的音素(如已经是平假名的)
if c_romaji is None or v_romaji is None:
continue
combined_romaji = c_romaji + v_romaji
# 检查组合是否合理(简单验证)
if v_romaji not in valid_vowels and len(v_romaji) > 2:
continue
# 转换为最终别名格式
if use_hiragana:
final_alias = ROMAJI_TO_HIRAGANA.get(combined_romaji)
# 如果无法转换为平假名,跳过此组合
if final_alias is None:
continue
else:
final_alias = combined_romaji
# 检查是否已存在(检查最终别名)
if final_alias in existing_aliases:
continue
# 同时检查罗马音形式是否已存在
if combined_romaji in existing_aliases:
continue
candidates.append({
"alias": final_alias,
"base_alias": combined_romaji, # 始终使用罗马音作为基础
"consonant_info": c_info,
"vowel_info": v_info
})
# 模糊拼字:生成使用近似音素的额外候选
if fuzzy_phoneme and language in ('chinese', 'zh', 'mandarin'):
fuzzy_candidates = self._generate_fuzzy_candidates(
consonants, vowels,
available_consonants, available_vowels,
existing_aliases, candidates
)
candidates.extend(fuzzy_candidates)
return candidates
def _find_fuzzy_substitute(
self,
phoneme: str,
available_phonemes: set,
groups: List[Tuple[str, ...]]
) -> Optional[str]:
"""
查找模糊替代音素
参数:
phoneme: 目标音素
available_phonemes: 可用音素集合
groups: 近似音素组列表(同组内音素互为替代)
返回:
替代音素,如果无法替代则返回 None
"""
# 如果目标音素已存在,直接返回
if phoneme in available_phonemes:
return phoneme
# 查找目标音素所在的近似组
for group in groups:
if phoneme in group:
# 按组内顺序查找可用的替代音素
for candidate in group:
if candidate != phoneme and candidate in available_phonemes:
return candidate
# 该组内没有可用替代
break
return None
def _generate_fuzzy_candidates(
self,
consonants: Dict[str, Dict],
vowels: Dict[str, Dict],
available_consonants: set,
available_vowels: set,
existing_aliases: set,
normal_candidates: List[Dict]
) -> List[Dict]:
"""
生成模糊拼字候选
使用近似音素替代缺失的声母/韵母,生成额外的候选组合
"""
fuzzy_candidates = []
# 已生成的别名(包括普通候选)
generated_aliases = set(c["base_alias"] for c in normal_candidates)
generated_aliases.update(existing_aliases)
# 中文所有可能的声母
all_consonants = ['b', 'p', 'm', 'f', 'd', 't', 'n', 'l', 'g', 'k', 'h',
'j', 'q', 'x', 'zh', 'ch', 'sh', 'r', 'z', 'c', 's', 'y', 'w']
# 中文所有可能的韵母(包含所有标准韵母)
all_vowels = ['a', 'o', 'e', 'i', 'u', 'v',
'ai', 'ei', 'ao', 'ou',
'an', 'en', 'ang', 'eng', 'ong',
'ia', 'ie', 'iao', 'iu', 'ian', 'in', 'iang', 'ing', 'iong',
'ua', 'uo', 'uai', 'ui', 'uan', 'un', 'uang', 'ueng',
've', 'van', 'vn', 'er']
fuzzy_count = 0
for target_c in all_consonants:
for target_v in all_vowels:
target_alias = target_c + target_v
# 跳过已存在或已生成的
if target_alias in generated_aliases:
continue
# 确定实际使用的辅音
if target_c in available_consonants:
actual_c = target_c
else:
actual_c = self._find_fuzzy_substitute(
target_c, available_consonants, FUZZY_CONSONANT_GROUPS
)
# 确定实际使用的元音
if target_v in available_vowels:
actual_v = target_v
else:
actual_v = self._find_fuzzy_substitute(
target_v, available_vowels, FUZZY_VOWEL_GROUPS
)
# 如果辅音或元音无法获取,跳过
if actual_c is None or actual_v is None:
continue
# 如果实际音素与目标相同,说明不需要模糊替换(普通候选已处理)
if actual_c == target_c and actual_v == target_v:
continue
# 获取音素信息
c_info = consonants.get(actual_c)
v_info = vowels.get(actual_v)
if c_info is None or v_info is None:
continue
fuzzy_candidates.append({
"alias": target_alias,
"base_alias": target_alias,
"consonant_info": c_info,
"vowel_info": v_info,
"is_fuzzy": True,
"fuzzy_from": f"{actual_c}+{actual_v}"
})
generated_aliases.add(target_alias)
fuzzy_count += 1
if fuzzy_count > 0:
self._log(f"模糊拼字生成 {fuzzy_count} 个额外候选")
return fuzzy_candidates
def _combine_and_save(
self,
candidate: Dict,
slices_dir: str,
export_dir: str,
overlap_ratio: float,
crossfade_ms: int,
first_naming_rule: str
) -> Tuple[Optional[Dict], Optional[str]]:
"""
执行音频拼接并保存
参数:
candidate: 候选信息
slices_dir: 切片目录
export_dir: 导出目录
overlap_ratio: overlap 比例
crossfade_ms: 交叉淡化时长
first_naming_rule: 命名规则
返回:
(oto条目, wav文件名) 或 (None, None)
"""
import numpy as np
import soundfile as sf
c_info = candidate["consonant_info"]
v_info = candidate["vowel_info"]
alias = candidate["alias"]
# 加载辅音片段
c_audio, c_sr = sf.read(c_info["wav_path"])
if len(c_audio.shape) > 1:
c_audio = c_audio.mean(axis=1)
c_start = int(c_info["offset_ms"] / 1000 * c_sr)
c_duration = int(c_info["duration_ms"] / 1000 * c_sr)
c_segment = c_audio[c_start:c_start + c_duration]
# 加载元音片段
v_audio, v_sr = sf.read(v_info["wav_path"])
if len(v_audio.shape) > 1:
v_audio = v_audio.mean(axis=1)
v_start = int(v_info["offset_ms"] / 1000 * v_sr)
v_duration = int(v_info["duration_ms"] / 1000 * v_sr)
v_segment = v_audio[v_start:v_start + v_duration]
# 确保采样率一致
if c_sr != v_sr:
logger.warning(f"采样率不一致: {c_sr} vs {v_sr},跳过")
return None, None
sr = c_sr
# 检查片段有效性
if len(c_segment) == 0 or len(v_segment) == 0:
return None, None
# 执行交叉淡化拼接
crossfade_samples = int(crossfade_ms / 1000 * sr)
crossfade_samples = min(crossfade_samples, len(c_segment) // 2, len(v_segment) // 2)
if crossfade_samples < 1:
crossfade_samples = 1
combined = self._crossfade_concat(c_segment, v_segment, crossfade_samples)
# 生成文件名(使用 C 前缀表示 Combined)
wav_name = f"C{candidate['alias']}.wav"
wav_path = os.path.join(export_dir, wav_name)
# 保存音频
sf.write(wav_path, combined, sr)
# 计算 oto 参数
c_duration_ms = c_info["duration_ms"]
total_duration_ms = len(combined) / sr * 1000
# 应用命名规则(作为首个样本)
final_alias = self.apply_naming_rule(first_naming_rule, alias, 0) if first_naming_rule else alias
entry = {
"wav_name": wav_name,
"alias": final_alias,
"offset": 0,
"consonant": round(c_duration_ms, 1),
"cutoff": round(-total_duration_ms, 1),
"preutterance": round(c_duration_ms, 1),
"overlap": round(c_duration_ms * overlap_ratio, 1),
"segment_duration": total_duration_ms,
"is_combined": True # 标记为拼接生成
}
return entry, wav_name
def _crossfade_concat(
self,
audio1: 'np.ndarray',
audio2: 'np.ndarray',
crossfade_samples: int
) -> 'np.ndarray':
"""
交叉淡化拼接两段音频
参数:
audio1: 第一段音频
audio2: 第二段音频
crossfade_samples: 交叉淡化采样数
返回:
拼接后的音频
"""
import numpy as np
if crossfade_samples <= 0:
return np.concatenate([audio1, audio2])
# 确保交叉淡化长度不超过音频长度
crossfade_samples = min(crossfade_samples, len(audio1), len(audio2))
# 创建淡入淡出曲线
fade_out = np.linspace(1.0, 0.0, crossfade_samples)
fade_in = np.linspace(0.0, 1.0, crossfade_samples)
# 分离各部分
part1 = audio1[:-crossfade_samples]
overlap1 = audio1[-crossfade_samples:]
overlap2 = audio2[:crossfade_samples]
part2 = audio2[crossfade_samples:]
# 交叉混合
crossfaded = overlap1 * fade_out + overlap2 * fade_in
# 拼接
return np.concatenate([part1, crossfaded, part2])
|