update vocaloid phoneme parsers

libresvip/plugins/svip/binsvip_generator.py

@@ -175,13 +175,12 @@ def generate_note(self, note: Note) -> XSNote:
                     value=svip_note_head_tags.get(note.head_tag, XSNoteHeadTagEnum.NoTag)
                 ),
                 lyric=note.lyric or DEFAULT_CHINESE_LYRIC,
+                pronouncing=note.pronunciation or "",
             )
             xs_note.width_pos = (
                 round(self.synchronizer.get_actual_ticks_from_ticks(note.end_pos))
                 - xs_note.start_pos
             )
-            if note.pronunciation:
-                xs_note.pronouncing = note.pronunciation
             if note.edited_phones is not None:
                 xs_note.note_phone_info = self.generate_phones(note.edited_phones)
             if note.vibrato is not None:

libresvip/plugins/vpr/vpr_parser.py

@@ -140,7 +140,7 @@ def parse_notes(self, notes: list[VocaloidNotes], pos: int, default_lyric: str)
                 length=note.duration,
                 key_number=note.number,
                 lyric=note.lyric or default_lyric,
-                pronunciation=note.phoneme,
+                pronunciation=None,
             )
             for note in notes
         ]

libresvip/plugins/vsq/options.py

@@ -1,12 +1,24 @@
+from enum import Enum
+from typing import Annotated
+
 from pydantic import BaseModel, Field
 
 
+class BreathOption(Enum):
+    IGNORE: Annotated[str, Field(title="Ignore all breath notes")] = "ignore"
+    KEEP: Annotated[str, Field(title="Keep as normal notes")] = "keep"
+
+
 class InputOptions(BaseModel):
     lyric_encoding: str = Field(
         default="shift-jis",
         title="Lyric text encoding",
         description="Unless the lyrics are garbled, this option should not be changed.",
     )
+    breath: BreathOption = Field(
+        default=BreathOption.IGNORE,
+        title="The way to handle breath notes",
+    )
 
 
 class OutputOptions(BaseModel):

libresvip/plugins/vsq/vsq_parser.py

@@ -1,6 +1,7 @@
 import configparser
 import dataclasses
 import math
+import re
 from typing import Optional
 
 import mido_fix as mido
@@ -16,13 +17,15 @@
     TimeSignature,
 )
 
-from .options import InputOptions
+from .options import BreathOption, InputOptions
 from .vocaloid_pitch import (
     ControllerEvent,
     VocaloidPartPitchData,
     pitch_from_vocaloid_parts,
 )
 
+BREATH_PATTERN = re.compile("br[1-5]")
+
 
 @dataclasses.dataclass
 class VsqParser:
@@ -188,16 +191,21 @@ def parse_notes(self, vsq_track: configparser.ConfigParser, tick_prefix: int) ->
                 if vsq_note["type"] == "Anote":
                     if (length := vsq_note.getint("length")) and (key := vsq_note.getint("note#")):
                         lyric_handle = vsq_note.get("lyrichandle", fallback="")
-                        lyric_value, xsampa_value = vsq_track.get(
+                        lyric_value, phoneme_value = vsq_track.get(
                             lyric_handle, "L0", fallback=","
                         ).split(",")[:2]
+                        if (
+                            self.options.breath == BreathOption.IGNORE
+                            and BREATH_PATTERN.fullmatch(phoneme_value.strip('"')) is not None
+                        ):
+                            continue
                         notes.append(
                             Note(
                                 start_pos=tick,
                                 length=length,
                                 key_number=key,
                                 lyric=lyric_value.strip('"'),
-                                pronunciation=xsampa_value.strip('"'),
+                                pronunciation=None,
                             )
                         )
         return notes

libresvip/plugins/vsqx/vsqx_parser.py

@@ -126,7 +126,7 @@ def parse_notes(self, notes: list[VsqxNote], tick_offset: int) -> list[Note]:
                 start_pos=note.pos_tick + tick_offset,
                 length=note.dur_tick,
                 lyric=note.lyric or DEFAULT_ENGLISH_LYRIC,
-                pronunciation=note.phnms.value if note.phnms is not None else None,
+                pronunciation=None,
                 key_number=note.note_num,
             )
             for note in notes