fix(ustx, seqtool, gui): fix in-place editing, interpolation, and uni…

README.en.md

@@ -83,6 +83,13 @@ When using a DiffSinger virtual singer for covers, users often already have an O
 
 A new USTX file with expression parameters added. The original project will not be modified.
 
+> [!TIP]
+> Starting from `v0.9.1`, if you prefer not to generate a new project file, you can **set the output path to be the same as the input project path**. In this case, the expression parameters will be written directly into the original project file.
+>
+> Under normal circumstances, the program will only update the specified expression parameters in the selected track. It will not affect other parameters or modify other tracks.
+>
+> ⚠️ **If you plan to use this feature, please make a backup in advance to prevent potential data loss in case of errors**.
+
 ## ✨ Features
 
 * [x] Windows support

README.md

@@ -83,6 +83,13 @@
 
 一个携带表情参数的新 USTX 文件。原始工程不会被修改。
 
+> [!TIP]
+> 从 `v0.9.1` 开始，如果您不希望额外生成新的工程文件，可以**将输出路径设置为与输入工程路径一致**。这样，表情参数会直接写入原始工程文件中。
+>
+> 正常情况下，程序只会更新您所选音轨中的指定表情参数，不会影响其它参数，也不会修改其他音轨。
+>
+> ⚠️ **如果您需要使用此功能，请提前做好备份，以防程序出错**。
+
 ## ✨ 功能特性
 
 * [x] Windows 支持

expressions/dyn.py

@@ -25,7 +25,7 @@ class DynLoader(ExpressionLoader):
     expression_name = "dyn"
     expression_info = _l("Dynamics (curve)")
     args = SimpleNamespace(
-        trim_silence     = Args(name="trim_silence"    , type=bool , default=True, help=_l("**Trim silence** from the leading and trailing edges of the audio before extracting expression")),  # noqa: E501
+        trim_silence     = Args(name="trim_silence"    , type=bool , default=True, help=_l("**Trim silence** from the leading and trailing edges of the audio before extracting expression\n\n**NOTICE**: This may slightly cut into the beginning and ending of voiced segments. If the effect is too severe, consider disabling this option\n\n")),  # noqa: E501
         align_radius     = Args(name="align_radius"    , type=int  , default=1   , help=_l("**Radius** for the FastDTW alignment algorithm; larger values allow more flexible alignment but increase computation time")),  # noqa: E501
         smoothness       = Args(name="smoothness"      , type=int  , default=2   , help=_l("Controls the **smoothness** of the expression curve using Gaussian filtering. Higher values produce smoother curves but may lose fine detail")),  # noqa: E501
         scaler           = Args(name="scaler"          , type=float, default=1.5 , help=_l("**Scaling factor** applied to the expression curve. Values >1 amplify the expression, =1 keeps original intensity, <1 reduces it")),  # noqa: E501

expressions/pitd.py

@@ -75,7 +75,6 @@ def get_expression(
             utau_time, utau_pitch, utau_confidence, utau_features = get_wav_features(
                 wav_path=self.utau_path, confidence_threshold=confidence_utau, backend=backend
             )
-        with StreamToLogger(self.logger, tee=True):
             ref_time, ref_pitch, ref_confidence, ref_features = get_wav_features(
                 wav_path=self.ref_path, confidence_threshold=confidence_ref, backend=backend
             )

expressions/tenc.py

@@ -25,7 +25,7 @@ class TencLoader(ExpressionLoader):
     expression_name = "tenc"
     expression_info = _l("Tension (curve)")
     args = SimpleNamespace(
-        trim_silence     = Args(name="trim_silence"    , type=bool , default=True, help=_l("**Trim silence** from the leading and trailing edges of the audio before extracting expression")),  # noqa: E501
+        trim_silence     = Args(name="trim_silence"    , type=bool , default=True, help=_l("**Trim silence** from the leading and trailing edges of the audio before extracting expression\n\n**NOTICE**: This may slightly cut into the beginning and ending of voiced segments. If the effect is too severe, consider disabling this option\n\n")),  # noqa: E501
         align_radius     = Args(name="align_radius"    , type=int  , default=1   , help=_l("**Radius** for the FastDTW alignment algorithm; larger values allow more flexible alignment but increase computation time")),  # noqa: E501
         smoothness       = Args(name="smoothness"      , type=int  , default=6   , help=_l("Controls the **smoothness** of the expression curve using Gaussian filtering. Higher values produce smoother curves but may lose fine detail")),  # noqa: E501
         scaler           = Args(name="scaler"          , type=float, default=1.0 , help=_l("**Scaling factor** applied to the expression curve. Values >1 amplify the expression, =1 keeps original intensity, <1 reduces it")),  # noqa: E501

expressive.py

@@ -1,10 +1,10 @@
 import os
 import logging
 import argparse
-from shutil import copy
 from pathlib import Path
 from datetime import datetime
 from contextlib import contextmanager
+from shutil import copy, SameFileError
 from os.path import splitext, basename
 
 from utils.i18n import init_gettext, _
@@ -77,7 +77,10 @@ def process_expressions(
 ]
 ```
     """
-    copy(ustx_input, ustx_output)
+    try:
+        copy(ustx_input, ustx_output)
+    except SameFileError:
+        pass
 
     for exp in expressions:
         exp_type = exp["expression"]

expressive_gui.py

@@ -161,7 +161,7 @@ async def export_config(state=state):
         if file and len(file) > 0:
             try:
                 with open(file[0], "w+", encoding="utf-8-sig") as f:  # type: ignore
-                    json.dump(state, f, indent=4)
+                    json.dump(state, f, indent=4, ensure_ascii=False)
                 ui.notify(_("Config exported successfully!"), type="positive")
             except Exception as e:
                 ui.notify(_("Failed to export config") + f": {str(e)}", type="negative")

locales/app.pot

@@ -290,7 +290,11 @@ msgstr ""
 #: expressions/dyn.py:28 expressions/tenc.py:28
 msgid ""
 "**Trim silence** from the leading and trailing edges of the audio before "
-"extracting expression"
+"extracting expression\n"
+"\n"
+"**NOTICE**: This may slightly cut into the beginning and ending of voiced"
+" segments. If the effect is too severe, consider disabling this option\n"
+"\n"
 msgstr ""
 
 #: expressions/dyn.py:29 expressions/pitd.py:41 expressions/tenc.py:29
@@ -474,15 +478,19 @@ msgid ""
 "curve upward; negative values shift it downward"
 msgstr ""
 
-#: utils/ui.py:353
+#: utils/ui.py:350
+msgid "Failed to load audio"
+msgstr ""
+
+#: utils/ui.py:377
 msgid "Play/Pause"
 msgstr ""
 
-#: utils/ui.py:354
+#: utils/ui.py:378
 msgid "Loop region"
 msgstr ""
 
-#: utils/ui.py:355
+#: utils/ui.py:379
 msgid "Zoom"
 msgstr ""
 

locales/en/LC_MESSAGES/app.po

@@ -301,10 +301,18 @@ msgstr "Dynamics (curve)"
 #: expressions/dyn.py:28 expressions/tenc.py:28
 msgid ""
 "**Trim silence** from the leading and trailing edges of the audio before "
-"extracting expression"
+"extracting expression\n"
+"\n"
+"**NOTICE**: This may slightly cut into the beginning and ending of voiced"
+" segments. If the effect is too severe, consider disabling this option\n"
+"\n"
 msgstr ""
 "**Trim silence** from the leading and trailing edges of the audio before "
-"extracting expression"
+"extracting expression\n"
+"\n"
+"**NOTICE**: This may slightly cut into the beginning and ending of voiced"
+" segments. If the effect is too severe, consider disabling this option\n"
+"\n"
 
 #: expressions/dyn.py:29 expressions/pitd.py:41 expressions/tenc.py:29
 msgid ""
@@ -517,15 +525,19 @@ msgstr ""
 "**Bias** offset added to the expression curve. Positive values shift the "
 "curve upward; negative values shift it downward"
 
-#: utils/ui.py:353
+#: utils/ui.py:350
+msgid "Failed to load audio"
+msgstr "Failed to load audio"
+
+#: utils/ui.py:377
 msgid "Play/Pause"
 msgstr "Play/Pause"
 
-#: utils/ui.py:354
+#: utils/ui.py:378
 msgid "Loop region"
 msgstr "Loop region"
 
-#: utils/ui.py:355
+#: utils/ui.py:379
 msgid "Zoom"
 msgstr "Zoom"
 

locales/zh_CN/LC_MESSAGES/app.po

@@ -291,8 +291,16 @@ msgstr "动态曲线 Dynamics (curve)"
 #: expressions/dyn.py:28 expressions/tenc.py:28
 msgid ""
 "**Trim silence** from the leading and trailing edges of the audio before "
-"extracting expression"
-msgstr "在提取表情特征前，**剪除**音频开头与结尾的**静音部分**"
+"extracting expression\n"
+"\n"
+"**NOTICE**: This may slightly cut into the beginning and ending of voiced"
+" segments. If the effect is too severe, consider disabling this option\n"
+"\n"
+msgstr ""
+"在提取表情特征前，**剪除**音频开头与结尾的**静音部分**\n"
+"\n"
+"**注意**：该操作可能会轻微截断有声段的起始和结束部分；若影响较为明显，建议关闭此功能\n"
+"\n"
 
 #: expressions/dyn.py:29 expressions/pitd.py:41 expressions/tenc.py:29
 msgid ""
@@ -487,15 +495,19 @@ msgid ""
 "curve upward; negative values shift it downward"
 msgstr "添加到表情曲线的**偏置**偏移量；正值使曲线上移，负值使曲线下移"
 
-#: utils/ui.py:353
+#: utils/ui.py:350
+msgid "Failed to load audio"
+msgstr "音频加载失败"
+
+#: utils/ui.py:377
 msgid "Play/Pause"
 msgstr "播放/暂停"
 
-#: utils/ui.py:354
+#: utils/ui.py:378
 msgid "Loop region"
 msgstr "选区循环播放"
 
-#: utils/ui.py:355
+#: utils/ui.py:379
 msgid "Zoom"
 msgstr "缩放"
 

tests/test_ustx.py

@@ -25,7 +25,6 @@
     TimeAxis,
     UstxEditor,
     RESOLUTION,
-    SUPPORTED_EXPRESSIONS,
 )
 
 
@@ -295,23 +294,6 @@ def test_set_curve_overwrites(self):
         assert len(part.curves) == 1
         assert part.get_curve("dyn").xs == [0, 960]
 
-    @pytest.mark.parametrize("abbr", sorted(SUPPORTED_EXPRESSIONS))
-    def test_set_curve_supported_expressions(self, abbr):
-        part = UVoicePart(track_no=0, position=0, duration=960)
-        part.set_curve(abbr, np.array([0, 480]), np.array([0.0, 50.0]))
-        assert part.get_curve(abbr) is not None
-
-    def test_set_curve_unsupported_expression(self):
-        part = UVoicePart(track_no=0, position=0, duration=960)
-        with pytest.raises(ValueError, match="Unsupported expression"):
-            part.set_curve("invalid_expr", np.array([0]), np.array([0.0]))
-
-    def test_set_curve_multiple_expressions(self):
-        part = UVoicePart(track_no=0, position=0, duration=960)
-        for abbr in SUPPORTED_EXPRESSIONS:
-            part.set_curve(abbr, np.array([0, 480]), np.array([0.0, 50.0]))
-        assert len(part.curves) == len(SUPPORTED_EXPRESSIONS)
-
 
 # ===========================================================================
 # TimeAxis
@@ -709,18 +691,6 @@ def test_full_workflow_via_editor(self, temp_ustx_file):
         assert curve is not None
         assert len(curve.xs) == 3
 
-    def test_multiple_expressions_persist(self, temp_ustx_file):
-        with UstxEditor(str(temp_ustx_file)) as editor:
-            ticks = np.array([0, 480, 960])
-            for abbr in sorted(SUPPORTED_EXPRESSIONS):
-                editor.add_expression_to_part(
-                    editor.voice_parts[0], abbr, ticks, np.array([1.0, 2.0, 3.0])
-                )
-
-        final = load_ustx(str(temp_ustx_file))
-        for abbr in SUPPORTED_EXPRESSIONS:
-            assert final.voice_parts[0].get_curve(abbr) is not None
-
     def test_time_axis_used_for_ticks(self, temp_ustx_file):
         """Verify ticks produced by TimeAxis match expected values."""
         with UstxEditor(str(temp_ustx_file)) as editor:

utils/seqtool.py

@@ -122,15 +122,15 @@ def unify_sequence_time(seq_times, seq_vals, to_ticks=False):
     if not to_ticks:
         unified_seq_time = np.unique(unified_seq_time)
         unified_seqs_val = [
-            interp1d(st, sv, fill_value="extrapolate")(unified_seq_time)  # type: ignore
+            interp1d(st, sv, fill_value=np.nan, bounds_error=False)(unified_seq_time)  # type: ignore
             for (st, sv) in zip(seq_times, seq_vals, strict=False)
         ]
         return unified_seq_time, tuple(unified_seqs_val)
 
     unified_seq_ticks = np.unique(_time_to_ticks_fn(unified_seq_time))
     time_mapping = _ticks_to_time_fn(unified_seq_ticks)
     unified_seqs_val = [
-        interp1d(st, sv, fill_value="extrapolate")(time_mapping)  # type: ignore
+        interp1d(st, sv, fill_value=np.nan, bounds_error=False)(time_mapping)  # type: ignore
         for (st, sv) in zip(seq_times, seq_vals, strict=False)
     ]
     return unified_seq_ticks, tuple(unified_seqs_val)
@@ -262,7 +262,7 @@ def align_sequence_tick(
     for q in unified_queries:
         aligned_tick = np.interp(path[:, 1], np.arange(len(unified_tick)), unified_tick)
         aligned_seq  = np.interp(path[:, 0], np.arange(len(q)), q)
-        interp_seq   = interp1d(aligned_tick, aligned_seq, fill_value="extrapolate")  # type: ignore
+        interp_seq   = interp1d(aligned_tick, aligned_seq, fill_value=np.nan, bounds_error=False)  # type: ignore
         aligned_queries.append(interp_seq(unified_tick))
 
     return unified_tick, tuple(aligned_queries), tuple(unified_references)

utils/ui.py

@@ -327,6 +327,29 @@ def _build(self) -> None:
             waveform_div.classes("w-full rounded-lg overflow-hidden")
             waveform_div.style(f"min-height:{self._height}px; background:var(--ws-bg);")
 
+            self._error = (
+                ui.label('')
+                .classes('w-full text-xs text-red-400 hidden')
+            )
+
+            self._loader = (
+                ui.linear_progress(show_value=False)
+                .props('instant-feedback rounded indeterminate')
+                .classes('w-full hidden')
+                .style('height:3px; margin:0; color:#c800c8;')
+            )
+
+        ui.on(f'{iid}-loading', lambda e: (
+            self._error.set_visibility(False),
+            self._loader.set_visibility(True),
+        ))
+        ui.on(f'{iid}-ready', lambda e: self._loader.set_visibility(False))
+        ui.on(f'{iid}-error', lambda e: (
+            self._loader.set_visibility(False),
+            self._error.set_text(e.args.get('message') or _('Failed to load audio')),
+            self._error.set_visibility(True),
+        ))
+
         ws_opts: dict[str, Any] = {
             "container": f"#{iid}_waveform",
             "waveColor": self._wave_color,
@@ -371,6 +394,20 @@ def _build(self) -> None:
 
   window['{iid}'] = {{ ws, regions, loop: {loop_init} }};
 
+  ws.on('loading', (percent) => {{
+    emitEvent('{iid}-loading', {{ percent: percent }});
+  }});
+
+  ws.on('ready', () => {{
+    emitEvent('{iid}-ready', {{}});
+  }});
+
+  ws.on('error', (err) => {{
+    // Filter out non-fatal errors
+    if (ws.getDuration() > 0) return;
+    emitEvent('{iid}-error', {{ message: err?.message || '' }});
+  }});
+
   if ({show_controls_json}) {{
     // Overlay controls: inject a style block + overlay div into the waveform container
     const styleEl = document.createElement('style');
@@ -537,6 +574,9 @@ def serve_wav(wav_path: str) -> str:
     Each unique directory is registered once under /wav/<hashed_dir>.
     WaveSurfer then streams the file normally — no base64 overhead.
     """
+    if not wav_path or not os.path.exists(wav_path):
+        return ''
+
     import hashlib
     directory = os.path.dirname(os.path.abspath(wav_path))
     dir_hash  = hashlib.md5(directory.encode()).hexdigest()[:8]

utils/ustx.py

@@ -40,8 +40,6 @@
 RESOLUTION = 480          # pulses per quarter note — hardcoded in UProject.cs
 MS_PER_MIN = 60_000.0     # milliseconds per minute
 
-SUPPORTED_EXPRESSIONS = frozenset({"dyn", "pitd", "tenc"})
-
 
 # ---------------------------------------------------------------------------
 # Data structures
@@ -221,11 +219,6 @@ def set_curve(
             ticks:  1-D integer array of tick positions.
             values: 1-D float array of curve values; NaN entries are dropped.
         """
-        if abbr not in SUPPORTED_EXPRESSIONS:
-            raise ValueError(
-                f"Unsupported expression '{abbr}'. "
-                f"Supported: {sorted(SUPPORTED_EXPRESSIONS)}"
-            )
         mask = ~np.isnan(values)
         curve = self.get_or_create_curve(abbr)
         curve.xs = ticks[mask].astype(int).tolist()
@@ -566,7 +559,7 @@ def save_ustx(project: UProject, ustx_path: str) -> None:
         project:   The project to save.
         ustx_path: Destination path.
     """
-    output = oyaml.dump(project.to_dict(), Dumper=oyaml.Dumper)
+    output = oyaml.dump(project.to_dict(), Dumper=oyaml.Dumper, allow_unicode=True)
     with open(ustx_path, "w+", encoding="utf-8-sig") as fh:
         fh.write(output)
     log.debug("Saved USTX to %s", ustx_path)
@@ -705,6 +698,7 @@ def add_expression_to_track(
         if not parts:
             raise ValueError(f"No voice parts found for track_no {track_no}.")
 
+        curve_set = False
         ticks = np.asarray(expression_ticks, dtype=int)
         values = np.asarray(expression_values, dtype=float)
 
@@ -716,3 +710,11 @@ def add_expression_to_track(
                 continue
             relative_ticks = ticks[mask] - part_start
             part.set_curve(expression_name, relative_ticks, values[mask])
+            curve_set = True
+
+        if not curve_set:
+            log.warning(
+                "No expression points fit inside any part on track %d (0-based) for curve %s.",
+                track_no,
+                expression_name,
+            )