Spaces:
Running
Running
cockolo terada
commited on
Update gradio_tabs/single.py
Browse files- gradio_tabs/single.py +115 -172
gradio_tabs/single.py
CHANGED
@@ -27,6 +27,9 @@ from typing import Dict, Any, List, Tuple, Optional, Set
|
|
27 |
# --- タイムゾーン定義 ---
|
28 |
# グローバルな定数としてJSTを定義
|
29 |
JST = timezone(timedelta(hours=9), 'JST')
|
|
|
|
|
|
|
30 |
|
31 |
|
32 |
# --- モック(本来はライブラリからインポート) ---
|
@@ -72,14 +75,6 @@ class TTSModelHolder:
|
|
72 |
}
|
73 |
with open(model2_path / "style_settings.json", "w", encoding="utf-8") as f:
|
74 |
json.dump(style_settings_data, f, indent=2, ensure_ascii=False)
|
75 |
-
|
76 |
-
# Sample Merged Model
|
77 |
-
merged_model_path = p / "miku_90p_rinu_10p"
|
78 |
-
merged_model_path.mkdir(parents=True, exist_ok=True)
|
79 |
-
(merged_model_path / "G_merged.safetensors").touch()
|
80 |
-
with open(merged_model_path / "config.json", "w", encoding="utf-8") as f:
|
81 |
-
json.dump(config1, f, indent=2)
|
82 |
-
|
83 |
|
84 |
def refresh(self) -> List[str]:
|
85 |
"""
|
@@ -87,7 +82,6 @@ class TTSModelHolder:
|
|
87 |
更新後のモデルリストを返す。
|
88 |
"""
|
89 |
if self.root_dir.is_dir():
|
90 |
-
# is_dir()はシンボリックリンクされたディレクトリもTrueを返す
|
91 |
self.model_names = sorted([d.name for d in self.root_dir.iterdir() if d.is_dir()])
|
92 |
print(f"TTSModelHolder model list refreshed. Known models: {self.model_names}")
|
93 |
else:
|
@@ -101,7 +95,7 @@ class TTSModelHolder:
|
|
101 |
error_msg = (
|
102 |
f"Model '{model_name}' is not in the known list of TTSModelHolder. "
|
103 |
f"Current list: {self.model_names}. "
|
104 |
-
"Please refresh the model list by clicking the refresh button."
|
105 |
)
|
106 |
print(f"[ERROR] {error_msg}")
|
107 |
raise ValueError(error_msg)
|
@@ -141,10 +135,8 @@ DEFAULT_STYLE_WEIGHT=1.0
|
|
141 |
DEFAULT_WORKBENCH_PAUSE = 250
|
142 |
# ------------------------------------------------
|
143 |
|
144 |
-
|
145 |
-
OUTPUT_SIZE_LIMIT_GB = 5 # outputフォルダの容量上限をGB単位で指定
|
146 |
OUTPUT_SIZE_LIMIT_BYTES = OUTPUT_SIZE_LIMIT_GB * 1024**3
|
147 |
-
# ▲▲▲ 変更・追加点 1 ▲▲▲
|
148 |
|
149 |
|
150 |
# --- ヘルパー関数 ---
|
@@ -190,6 +182,16 @@ def format_and_sort_model_names(dir_list: List[str]) -> List[Tuple[str, str]]:
|
|
190 |
result_list.extend(sorted(unparsed_models))
|
191 |
return result_list
|
192 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
193 |
def set_random_seed(seed: int):
|
194 |
if seed >= 0:
|
195 |
print(f"Setting random seed to: {seed}")
|
@@ -200,7 +202,6 @@ def set_random_seed(seed: int):
|
|
200 |
np.random.seed(seed)
|
201 |
random.seed(seed)
|
202 |
|
203 |
-
# ▼▼▼ 変更・追加点 2: 容量計算用のヘルパー関数を追加 ▼▼▼
|
204 |
def get_directory_size(directory_path: Path) -> int:
|
205 |
"""指定されたディレクトリの合計サイズをバイト単位で返す。"""
|
206 |
total_size = 0
|
@@ -228,7 +229,6 @@ def format_bytes(size_bytes: int) -> str:
|
|
228 |
p = math.pow(1024, i)
|
229 |
s = round(size_bytes / p, 2)
|
230 |
return f"{s} {size_name[i]}"
|
231 |
-
# ▲▲▲ 変更・追加点 2 ▲▲▲
|
232 |
|
233 |
|
234 |
# --- pyopenjtalk関連ヘルパー関数 ---
|
@@ -422,15 +422,7 @@ def process_single_synthesis_webui(
|
|
422 |
return True, log_messages, (sr, audio_data)
|
423 |
|
424 |
|
425 |
-
# ▼▼▼ 変更点: action_save_audio 関数を削除 ▼▼▼
|
426 |
-
# この関数は不要になったため削除されました。
|
427 |
-
# ▲▲▲ 変更点 ▲▲▲
|
428 |
-
|
429 |
def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
430 |
-
# ▼▼▼ 変更点: is_shm_available 変数を削除 ▼▼▼
|
431 |
-
# is_shm_available = sys.platform != "win32" and Path("/dev/shm").exists() and Path("/dev/shm").is_dir()
|
432 |
-
# ▲▲▲ 変更点 ▲▲▲
|
433 |
-
|
434 |
custom_css = """
|
435 |
.audio-output-row { display: flex !important; flex-wrap: wrap !important; gap: 10px !important; }
|
436 |
.audio-item-column { flex-grow: 0 !important; flex-shrink: 0 !important; width: var(--audio-width, 250px) !important; background-color: #f8f9fa; padding: 8px; border-radius: 8px; border: 1px solid #dee2e6; }
|
@@ -445,18 +437,13 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
445 |
with gr.Blocks(css=custom_css) as app:
|
446 |
MAX_AUDIO_OUTPUTS = 4
|
447 |
ITEMS_PER_ROW = 4
|
448 |
-
# ▼▼▼ 変更点: キープの最大アイテム数を8に変更 ▼▼▼
|
449 |
MAX_WORKBENCH_ITEMS = 8
|
450 |
-
# ▲▲▲ 変更点 ▲▲▲
|
451 |
|
452 |
all_styles_data_state = gr.State({})
|
453 |
workbench_state = gr.State([])
|
454 |
merged_preview_state = gr.State({})
|
455 |
-
# ▼▼▼ 変更点: saved_audio_hashes_stateを削除 ▼▼▼
|
456 |
-
# saved_audio_hashes_state = gr.State(set())
|
457 |
-
# ▲▲▲ 変更点 ▲▲▲
|
458 |
|
459 |
-
# ---
|
460 |
def update_workbench_ui(workbench_list: List[Dict]) -> Tuple:
|
461 |
updates = []
|
462 |
for i in range(MAX_WORKBENCH_ITEMS):
|
@@ -506,9 +493,6 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
506 |
with gr.Row(elem_classes="audio-output-row"):
|
507 |
audio_item_columns = []
|
508 |
audio_outputs = []
|
509 |
-
# ▼▼▼ 変更点: save_buttons を削除 ▼▼▼
|
510 |
-
# save_buttons = []
|
511 |
-
# ▲▲▲ 変更点 ▲▲▲
|
512 |
to_workbench_buttons = []
|
513 |
synthesized_text_states = []
|
514 |
dummy_audio_item_columns = []
|
@@ -521,10 +505,7 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
521 |
type="filepath", interactive=False
|
522 |
))
|
523 |
with gr.Row():
|
524 |
-
|
525 |
-
# save_buttons.append(gr.Button("💾 保存", scale=1))
|
526 |
-
to_workbench_buttons.append(gr.Button("🛠️ キープに追加", scale=2))
|
527 |
-
# ▲▲▲ 変更点 ▲▲▲
|
528 |
audio_item_columns.append(audio_col)
|
529 |
|
530 |
for i in range(ITEMS_PER_ROW - 1):
|
@@ -532,27 +513,16 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
532 |
pass
|
533 |
dummy_audio_item_columns.append(dummy_col)
|
534 |
|
535 |
-
# ▼▼▼ 変更点: 保存設定アコーディオン(ID入力欄)を削除 ▼▼▼
|
536 |
-
# with gr.Accordion("保存設定", open=True):
|
537 |
-
# metadata_textbox = gr.Textbox(
|
538 |
-
# label="ID",
|
539 |
-
# lines=1, placeholder="音声ファイルに埋め込むメモを入力してください(必須)",
|
540 |
-
# interactive=True
|
541 |
-
# )
|
542 |
-
# ▲▲▲ 変更点 ▲▲▲
|
543 |
with gr.Accordion("ステータス", open=True):
|
544 |
status_textbox = gr.Textbox(interactive=False, lines=5, max_lines=5, autoscroll=True, show_label=False, placeholder="ここにログが表示されます...")
|
545 |
|
546 |
with gr.Column(scale=1):
|
547 |
-
# ▼▼▼ 変更点: 「融☆合モデルを使う」チェックボックスを削除し、レイアウトを調整 ▼▼▼
|
548 |
with gr.Row():
|
549 |
-
|
|
|
|
|
550 |
refresh_model_list_button = gr.Button("再読込", scale=1)
|
551 |
-
|
552 |
-
|
553 |
-
# ▼▼▼ 変更点: モデルファイル選択ドロップダウンを削除 ▼▼▼
|
554 |
-
# selected_model_file_dropdown = gr.Dropdown(label="モデルファイル (.safetensors)", choices=[], value=None, interactive=True)
|
555 |
-
# ▲▲▲ 変更点 ▲▲▲
|
556 |
current_styles_dropdown = gr.Dropdown(label="スタイル", choices=[], type="value", interactive=True)
|
557 |
style_weight_for_synth_slider = gr.Slider(label="スタイル強度", minimum=0.0, maximum=20.0, value=DEFAULT_STYLE_WEIGHT, step=0.1, info="初期値は推奨強度", interactive=True)
|
558 |
batch_count_slider = gr.Slider(label="生成数", value=1, minimum=1, maximum=MAX_AUDIO_OUTPUTS, step=1, interactive=True)
|
@@ -579,23 +549,17 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
579 |
random_text_mode_slider = gr.Slider(label="分割の単位", minimum=1, maximum=4, value=1, step=1, info="1:形態素, 2:チャンク, 3:文節, 4:節", interactive=True)
|
580 |
random_text_ratio_textbox = gr.Textbox(label="カタカナ化の割合", value="0.5, 1", info="カンマ区切りで複数指定可。指定値からランダムに1つ使用。", interactive=True)
|
581 |
|
582 |
-
|
583 |
-
|
584 |
-
gr.Markdown("## キープ\n読み上げタブで生成した音声をここにストックし、結合や保存ができます。最大8個まで保持できます。")
|
585 |
|
586 |
workbench_items = []
|
587 |
all_workbench_ui_components = []
|
588 |
|
589 |
with gr.Row(variant="panel"):
|
590 |
-
# 左側の音声リストエリア(2列)
|
591 |
with gr.Column(scale=3):
|
592 |
with gr.Row():
|
593 |
-
# 左列 (アイテム 1-4)
|
594 |
left_workbench_col = gr.Column(scale=1)
|
595 |
-
# 右列 (アイテム 5-8)
|
596 |
right_workbench_col = gr.Column(scale=1)
|
597 |
-
|
598 |
-
# 右側の操作パネル
|
599 |
with gr.Column(scale=1):
|
600 |
with gr.Blocks():
|
601 |
with gr.Row():
|
@@ -612,38 +576,13 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
612 |
|
613 |
with gr.Row():
|
614 |
merge_preview_button = gr.Button("1.結合&プレビュー", variant="primary")
|
615 |
-
add_merged_to_workbench_button = gr.Button("
|
616 |
delete_originals_checkbox = gr.Checkbox(label="結合時に自動で元ファイルを削除", value=False, interactive=True)
|
617 |
|
618 |
preview_audio_player = gr.Audio(label="結合結果プレビュー", interactive=False, type="filepath")
|
619 |
|
620 |
-
# ▼▼▼ 変更点: キープの保存機能を削除 ▼▼▼
|
621 |
-
# gr.Markdown("---")
|
622 |
-
#
|
623 |
-
# with gr.Blocks():
|
624 |
-
# gr.Markdown("#### キープの音声を保存")
|
625 |
-
# with gr.Row():
|
626 |
-
# with gr.Column(scale=1):
|
627 |
-
# audio_to_save_num_input = gr.Number(label="保存する音声の番号", value=1, minimum=1, step=1, precision=0, interactive=True)
|
628 |
-
# with gr.Column(scale=3):
|
629 |
-
# creative_filename_input = gr.Textbox(
|
630 |
-
# label="保存する時のファイル名",
|
631 |
-
# placeholder=f"例: れいさな",
|
632 |
-
# interactive=True
|
633 |
-
# )
|
634 |
-
# workbench_metadata_textbox = gr.Textbox(
|
635 |
-
# label="ID (必須)",
|
636 |
-
# lines=1, placeholder="この音声に関するメモを入力してください",
|
637 |
-
# interactive=True
|
638 |
-
# )
|
639 |
-
# save_creative_button = gr.Button("💾 指定番号の音声を保存", variant="primary")
|
640 |
-
# ▲▲▲ 変更点 ▲▲▲
|
641 |
-
|
642 |
-
|
643 |
-
# 2列にアイテムを配置するロジック
|
644 |
ITEMS_PER_COLUMN = 4
|
645 |
for i in range(MAX_WORKBENCH_ITEMS):
|
646 |
-
# iの値に応じて、アイテムを追加する親のColumnを決定
|
647 |
parent_column = left_workbench_col if i < ITEMS_PER_COLUMN else right_workbench_col
|
648 |
|
649 |
with parent_column:
|
@@ -658,20 +597,13 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
658 |
"audio": audio, "info": info, "delete_btn": delete_btn
|
659 |
})
|
660 |
|
661 |
-
# all_workbench_ui_componentsの生成ロジックは変更なし
|
662 |
for item in workbench_items:
|
663 |
all_workbench_ui_components.extend([
|
664 |
item["row"], item["item_num_display"], item["audio"], item["info"]
|
665 |
])
|
666 |
-
# ▲▲▲ 変更点 2 ▲▲▲
|
667 |
|
668 |
|
669 |
# --- UIイベントハンドラ関数 ---
|
670 |
-
|
671 |
-
# ▼▼▼ 変更点: update_model_files_dropdown 関数を削除 ▼▼▼
|
672 |
-
# この関数はモデルファイル選択UIの削除に伴い不要になりました。
|
673 |
-
# ▲▲▲ 変更点 ▲▲▲
|
674 |
-
|
675 |
def load_styles_for_ui(selected_model_name: Optional[str]):
|
676 |
if not selected_model_name: return gr.update(choices=[], value=None), gr.update(value=DEFAULT_STYLE_WEIGHT), {}
|
677 |
model_path = assets_root_path / selected_model_name
|
@@ -687,29 +619,60 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
687 |
default_weight = styles_map[first_key].get("weight", DEFAULT_STYLE_WEIGHT)
|
688 |
return gr.update(choices=display_names, value=default_display_name), gr.update(value=default_weight), styles_map
|
689 |
|
690 |
-
# ▼▼▼ 変更点:
|
691 |
-
def action_refresh_model_list():
|
692 |
"""モデルリストを再読み込みし、UIとバックエンドの状態を同期させる。"""
|
693 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
694 |
model_holder.refresh()
|
695 |
|
696 |
-
|
697 |
-
|
698 |
-
|
699 |
-
|
700 |
-
|
701 |
-
|
702 |
-
|
703 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
704 |
|
705 |
-
return model_dropdown_update, style_dropdown_update, style_weight_update, styles_data_state_update
|
706 |
# ▲▲▲ 変更点 ▲▲▲
|
707 |
|
|
|
708 |
def on_model_select_change(selected_model_name: Optional[str]):
|
709 |
-
# ▼▼▼ 変更点: モデルファイル選択UIがなくなったため、関連処理を削除 ▼▼▼
|
710 |
style_dropdown_update, style_weight_update, styles_data_state_update = load_styles_for_ui(selected_model_name)
|
711 |
return style_dropdown_update, style_weight_update, styles_data_state_update
|
712 |
-
# ▲▲▲ 変更点 ▲▲▲
|
713 |
|
714 |
def on_style_dropdown_select(selected_display_name: Optional[str], styles_data: Dict[str, Any]):
|
715 |
if not selected_display_name or not styles_data: return gr.update(value=DEFAULT_STYLE_WEIGHT)
|
@@ -720,9 +683,6 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
720 |
|
721 |
def action_run_synthesis(
|
722 |
model_name: Optional[str],
|
723 |
-
# ▼▼▼ 変更点: selected_model_file 引数を削除 ▼▼▼
|
724 |
-
# selected_model_file: Optional[str],
|
725 |
-
# ▲▲▲ 変更点 ▲▲▲
|
726 |
style_display_name: Optional[str], style_weight_for_synth: float,
|
727 |
text: str, generation_mode: str, batch_count: int,
|
728 |
lang: str, seed: int, speaker: str, ref_audio: Optional[str],
|
@@ -769,17 +729,14 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
769 |
|
770 |
all_logs = []
|
771 |
|
772 |
-
# ▼▼▼ 変更点: モデルファイルを自動で選択するロジック ▼▼▼
|
773 |
model_path = assets_root_path / model_name
|
774 |
files = find_safetensors_files_webui(str(model_path))
|
775 |
if not files:
|
776 |
error_outputs[0] = f"❌ [エラー] モデルフォルダ '{model_name}' に .safetensors ファイルが見つかりません。"
|
777 |
return tuple(error_outputs)
|
778 |
|
779 |
-
# ソートされたリストの最初のファイルを使用する
|
780 |
actual_model_file_to_load = str(model_path / files[0])
|
781 |
all_logs.append(f"[自動選択] 使用モデルファイル: {files[0]}")
|
782 |
-
# ▲▲▲ 変更点 ▲▲▲
|
783 |
|
784 |
batch_count = int(batch_count)
|
785 |
if batch_count <= 0: batch_count = 1
|
@@ -897,7 +854,7 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
897 |
|
898 |
return tuple(final_outputs)
|
899 |
|
900 |
-
# ---
|
901 |
def add_to_workbench(
|
902 |
current_status: str,
|
903 |
current_workbench_list: List[Dict],
|
@@ -907,12 +864,12 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
907 |
safe_workbench_list = current_workbench_list or []
|
908 |
|
909 |
if not audio_path or not Path(audio_path).exists():
|
910 |
-
log_messages.append("⚠️ [
|
911 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
912 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
913 |
|
914 |
if any(item['audio_path'] == audio_path for item in safe_workbench_list):
|
915 |
-
log_messages.append("ℹ️
|
916 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
917 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
918 |
|
@@ -941,10 +898,10 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
941 |
path_to_delete.unlink()
|
942 |
except Exception as e:
|
943 |
print(f"Warning: Failed to delete old workbench audio file: {e}")
|
944 |
-
log_messages.append(f"ℹ️
|
945 |
|
946 |
ui_updates = update_workbench_ui(updated_list)
|
947 |
-
log_messages.append("✅
|
948 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
949 |
return (final_status, updated_list) + ui_updates
|
950 |
|
@@ -959,13 +916,13 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
959 |
path_to_delete = Path(item_to_remove['audio_path'])
|
960 |
if path_to_delete.exists() and str(path_to_delete.parent) == tempfile.gettempdir():
|
961 |
path_to_delete.unlink()
|
962 |
-
log_messages.append(f"✅
|
963 |
elif path_to_delete.exists():
|
964 |
-
log_messages.append(f"✅
|
965 |
else:
|
966 |
-
log_messages.append(f"✅
|
967 |
except Exception as e:
|
968 |
-
log_messages.append(f"⚠️
|
969 |
|
970 |
updated_list = [item for i, item in enumerate(safe_workbench_list) if i != index_to_remove]
|
971 |
ui_updates = update_workbench_ui(updated_list)
|
@@ -980,7 +937,7 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
980 |
):
|
981 |
log_messages = []
|
982 |
if not workbench_list:
|
983 |
-
log_messages.append("⚠️ [結合プレビュー警告]
|
984 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
985 |
return final_status, None, {}
|
986 |
|
@@ -1008,33 +965,21 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1008 |
pause_duration = int(pause_ms)
|
1009 |
|
1010 |
if pause_duration >= 0:
|
1011 |
-
# 正の値: 無音を挿入
|
1012 |
pause_segment = AudioSegment.silent(duration=pause_duration)
|
1013 |
combined_audio = segment1 + pause_segment + segment2
|
1014 |
log_messages.append(f"✅ 音声 #{first_audio_num} と #{second_audio_num} を {pause_duration}ms のポーズを挟んで結合しました。")
|
1015 |
else:
|
1016 |
-
# 負の値: そのまま重ねる(オーバーレイ)
|
1017 |
overlap_duration = abs(pause_duration)
|
1018 |
-
|
1019 |
-
# オーバーラップ長がどちらかの音声の長さを超えないように制限
|
1020 |
max_possible_overlap = min(len(segment1), len(segment2))
|
1021 |
if overlap_duration > max_possible_overlap:
|
1022 |
log_messages.append(f"ℹ️ オーバーラップ長({overlap_duration}ms)が可能な最大値({max_possible_overlap}ms)を超えるため、自動的に調整されました。")
|
1023 |
overlap_duration = max_possible_overlap
|
1024 |
|
1025 |
-
# 結合後の最終的な音声の長さを計算
|
1026 |
final_duration = len(segment1) + len(segment2) - overlap_duration
|
1027 |
-
|
1028 |
-
# 最終的な長さの無音キャンバスを作成
|
1029 |
combined_audio = AudioSegment.silent(duration=final_duration)
|
1030 |
-
|
1031 |
-
# 1番目の音声をキャンバスの先頭から重ねる
|
1032 |
combined_audio = combined_audio.overlay(segment1, position=0)
|
1033 |
-
|
1034 |
-
# 2番目の音声を、1番目の音声の末尾からオーバーラップする位置に重ねる
|
1035 |
overlay_position = len(segment1) - overlap_duration
|
1036 |
combined_audio = combined_audio.overlay(segment2, position=overlay_position)
|
1037 |
-
|
1038 |
log_messages.append(f"✅ 音声 #{first_audio_num} と #{second_audio_num} を {overlap_duration}ms 重ねて(オーバーレイして)結合しました。")
|
1039 |
|
1040 |
progress(1, desc="結合完了")
|
@@ -1065,10 +1010,6 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1065 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
1066 |
return final_status, str(temp_path), metadata
|
1067 |
|
1068 |
-
# ▼▼▼ 変更点: action_save_workbench_audio 関数を削除 ▼▼▼
|
1069 |
-
# この関数は保存機能の削除に伴い不要になりました。
|
1070 |
-
# ▲▲▲ 変更点 ▲▲▲
|
1071 |
-
|
1072 |
def action_add_merged_to_workbench(
|
1073 |
current_status: str,
|
1074 |
preview_data: Dict,
|
@@ -1080,13 +1021,13 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1080 |
log_messages = []
|
1081 |
safe_workbench_list = current_workbench_list or []
|
1082 |
if not preview_data or "audio_path" not in preview_data:
|
1083 |
-
log_messages.append("⚠️ [
|
1084 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
1085 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
1086 |
|
1087 |
src_path = Path(preview_data["audio_path"])
|
1088 |
if not src_path.exists():
|
1089 |
-
log_messages.append("⚠️ [
|
1090 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
1091 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
1092 |
|
@@ -1118,10 +1059,10 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1118 |
log_messages.append(f"⚠️ 元の音声ファイル削除中にエラー: {e}")
|
1119 |
|
1120 |
final_workbench_list = [new_merged_item] + remaining_list
|
1121 |
-
log_messages.append(f"✅
|
1122 |
else:
|
1123 |
final_workbench_list = [new_merged_item] + safe_workbench_list
|
1124 |
-
log_messages.append("✅
|
1125 |
|
1126 |
if len(final_workbench_list) > MAX_WORKBENCH_ITEMS:
|
1127 |
item_to_remove = final_workbench_list.pop(-1)
|
@@ -1131,7 +1072,7 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1131 |
path_to_delete.unlink()
|
1132 |
except Exception as e:
|
1133 |
print(f"Warning: Failed to delete old workbench audio file: {e}")
|
1134 |
-
log_messages.append(f"ℹ️
|
1135 |
|
1136 |
ui_updates = update_workbench_ui(final_workbench_list)
|
1137 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
@@ -1139,12 +1080,14 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1139 |
|
1140 |
|
1141 |
# --- イベントリスナー接続 ---
|
1142 |
-
# ▼▼▼ 変更点:
|
1143 |
-
refresh_model_list_button.click
|
1144 |
-
|
1145 |
-
|
1146 |
-
|
1147 |
-
|
|
|
|
|
1148 |
|
1149 |
selected_model_dropdown.change(on_model_select_change,
|
1150 |
inputs=[selected_model_dropdown],
|
@@ -1162,8 +1105,7 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1162 |
generate_button.click(
|
1163 |
fn=action_run_synthesis,
|
1164 |
inputs=[
|
1165 |
-
selected_model_dropdown,
|
1166 |
-
# use_symlink_mode_checkbox は削除
|
1167 |
current_styles_dropdown, style_weight_for_synth_slider,
|
1168 |
text_input, generation_mode_radio, batch_count_slider,
|
1169 |
language_dropdown, seed_input, speaker_name_textbox,
|
@@ -1178,9 +1120,6 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1178 |
)
|
1179 |
|
1180 |
for i in range(MAX_AUDIO_OUTPUTS):
|
1181 |
-
# ▼▼▼ 変更点: 保存ボタンのクリックイベントを削除 ▼▼▼
|
1182 |
-
# save_buttons[i].click(...)
|
1183 |
-
# ▲▲▲ 変更点 ▲▲▲
|
1184 |
to_workbench_buttons[i].click(
|
1185 |
fn=add_to_workbench,
|
1186 |
inputs=[
|
@@ -1209,10 +1148,6 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1209 |
outputs=[status_textbox, preview_audio_player, merged_preview_state]
|
1210 |
)
|
1211 |
|
1212 |
-
# ▼▼▼ 変更点: キープの保存ボタンのクリックイベントを削除 ▼▼▼
|
1213 |
-
# save_creative_button.click(...)
|
1214 |
-
# ▲▲▲ 変更点 ▲▲▲
|
1215 |
-
|
1216 |
add_merged_to_workbench_button.click(
|
1217 |
fn=action_add_merged_to_workbench,
|
1218 |
inputs=[
|
@@ -1228,11 +1163,9 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1228 |
|
1229 |
player_width_slider.release(lambda w: f"<script>document.documentElement.style.setProperty('--audio-width', '{w}px');</script>", inputs=[player_width_slider], outputs=[js_injector_html])
|
1230 |
|
1231 |
-
app.load(
|
1232 |
-
|
1233 |
-
|
1234 |
-
outputs=[selected_model_dropdown, current_styles_dropdown, style_weight_for_synth_slider, all_styles_data_state]
|
1235 |
-
)
|
1236 |
# ▲▲▲ 変更点 ▲▲▲
|
1237 |
return app
|
1238 |
|
@@ -1240,10 +1173,13 @@ def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
1240 |
if __name__ == "__main__":
|
1241 |
if Path("model_assets").exists(): shutil.rmtree("model_assets")
|
1242 |
|
1243 |
-
# ▼▼▼ 変更点:
|
1244 |
-
#
|
1245 |
-
|
1246 |
-
|
|
|
|
|
|
|
1247 |
mock_model_holder = TTSModelHolder()
|
1248 |
print(f"Initial models loaded by TTSModelHolder: {mock_model_holder.model_names}")
|
1249 |
app = create_synthesis_app(mock_model_holder)
|
@@ -1252,10 +1188,16 @@ if __name__ == "__main__":
|
|
1252 |
assets_dir_path.mkdir(exist_ok=True)
|
1253 |
allowed_paths = [str(assets_dir_path)]
|
1254 |
|
1255 |
-
#
|
1256 |
-
|
1257 |
-
|
1258 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
1259 |
|
1260 |
output_dir_path = Path("output").resolve()
|
1261 |
(output_dir_path / "normal").mkdir(exist_ok=True, parents=True)
|
@@ -1267,4 +1209,5 @@ if __name__ == "__main__":
|
|
1267 |
|
1268 |
print(f"Gradioに次のパスへのアクセスを許可します: {', '.join(allowed_paths)}")
|
1269 |
|
1270 |
-
app.launch(allowed_paths=allowed_paths)
|
|
|
|
27 |
# --- タイムゾーン定義 ---
|
28 |
# グローバルな定数としてJSTを定義
|
29 |
JST = timezone(timedelta(hours=9), 'JST')
|
30 |
+
# ▼▼▼ 変更点: FNモデル用のキャッシュパスを定義 ▼▼▼
|
31 |
+
FN_MODEL_CACHE_PATH = Path("/tmp/sbv2_merger_cache")
|
32 |
+
# ▲▲▲ 変更点 ▲▲▲
|
33 |
|
34 |
|
35 |
# --- モック(本来はライブラリからインポート) ---
|
|
|
75 |
}
|
76 |
with open(model2_path / "style_settings.json", "w", encoding="utf-8") as f:
|
77 |
json.dump(style_settings_data, f, indent=2, ensure_ascii=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
78 |
|
79 |
def refresh(self) -> List[str]:
|
80 |
"""
|
|
|
82 |
更新後のモデルリストを返す。
|
83 |
"""
|
84 |
if self.root_dir.is_dir():
|
|
|
85 |
self.model_names = sorted([d.name for d in self.root_dir.iterdir() if d.is_dir()])
|
86 |
print(f"TTSModelHolder model list refreshed. Known models: {self.model_names}")
|
87 |
else:
|
|
|
95 |
error_msg = (
|
96 |
f"Model '{model_name}' is not in the known list of TTSModelHolder. "
|
97 |
f"Current list: {self.model_names}. "
|
98 |
+
"Please refresh the model list by toggling the symlink checkbox or clicking the refresh button."
|
99 |
)
|
100 |
print(f"[ERROR] {error_msg}")
|
101 |
raise ValueError(error_msg)
|
|
|
135 |
DEFAULT_WORKBENCH_PAUSE = 250
|
136 |
# ------------------------------------------------
|
137 |
|
138 |
+
OUTPUT_SIZE_LIMIT_GB = 5
|
|
|
139 |
OUTPUT_SIZE_LIMIT_BYTES = OUTPUT_SIZE_LIMIT_GB * 1024**3
|
|
|
140 |
|
141 |
|
142 |
# --- ヘルパー関数 ---
|
|
|
182 |
result_list.extend(sorted(unparsed_models))
|
183 |
return result_list
|
184 |
|
185 |
+
# ▼▼▼ 変更点: FNモデルソート用のヘルパー関数を追加 ▼▼▼
|
186 |
+
def get_fn_model_sort_key(name: str) -> int:
|
187 |
+
"""FNモデル名からソート用の数値を抽出する。例: 'FN10' -> 10"""
|
188 |
+
match = re.search(r'FN(\d+)', name, re.IGNORECASE)
|
189 |
+
if match:
|
190 |
+
return int(match.group(1))
|
191 |
+
return float('inf') # マッチしないものは後ろに配置
|
192 |
+
# ▲▲▲ 変更点 ▲▲▲
|
193 |
+
|
194 |
+
|
195 |
def set_random_seed(seed: int):
|
196 |
if seed >= 0:
|
197 |
print(f"Setting random seed to: {seed}")
|
|
|
202 |
np.random.seed(seed)
|
203 |
random.seed(seed)
|
204 |
|
|
|
205 |
def get_directory_size(directory_path: Path) -> int:
|
206 |
"""指定されたディレクトリの合計サイズをバイト単位で返す。"""
|
207 |
total_size = 0
|
|
|
229 |
p = math.pow(1024, i)
|
230 |
s = round(size_bytes / p, 2)
|
231 |
return f"{s} {size_name[i]}"
|
|
|
232 |
|
233 |
|
234 |
# --- pyopenjtalk関連ヘルパー関数 ---
|
|
|
422 |
return True, log_messages, (sr, audio_data)
|
423 |
|
424 |
|
|
|
|
|
|
|
|
|
425 |
def create_synthesis_app(model_holder: TTSModelHolder) -> gr.Blocks:
|
|
|
|
|
|
|
|
|
426 |
custom_css = """
|
427 |
.audio-output-row { display: flex !important; flex-wrap: wrap !important; gap: 10px !important; }
|
428 |
.audio-item-column { flex-grow: 0 !important; flex-shrink: 0 !important; width: var(--audio-width, 250px) !important; background-color: #f8f9fa; padding: 8px; border-radius: 8px; border: 1px solid #dee2e6; }
|
|
|
437 |
with gr.Blocks(css=custom_css) as app:
|
438 |
MAX_AUDIO_OUTPUTS = 4
|
439 |
ITEMS_PER_ROW = 4
|
|
|
440 |
MAX_WORKBENCH_ITEMS = 8
|
|
|
441 |
|
442 |
all_styles_data_state = gr.State({})
|
443 |
workbench_state = gr.State([])
|
444 |
merged_preview_state = gr.State({})
|
|
|
|
|
|
|
445 |
|
446 |
+
# --- 作業台UI更新ヘルパー ---
|
447 |
def update_workbench_ui(workbench_list: List[Dict]) -> Tuple:
|
448 |
updates = []
|
449 |
for i in range(MAX_WORKBENCH_ITEMS):
|
|
|
493 |
with gr.Row(elem_classes="audio-output-row"):
|
494 |
audio_item_columns = []
|
495 |
audio_outputs = []
|
|
|
|
|
|
|
496 |
to_workbench_buttons = []
|
497 |
synthesized_text_states = []
|
498 |
dummy_audio_item_columns = []
|
|
|
505 |
type="filepath", interactive=False
|
506 |
))
|
507 |
with gr.Row():
|
508 |
+
to_workbench_buttons.append(gr.Button("🛠️ 作業台に追加", scale=2))
|
|
|
|
|
|
|
509 |
audio_item_columns.append(audio_col)
|
510 |
|
511 |
for i in range(ITEMS_PER_ROW - 1):
|
|
|
513 |
pass
|
514 |
dummy_audio_item_columns.append(dummy_col)
|
515 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
516 |
with gr.Accordion("ステータス", open=True):
|
517 |
status_textbox = gr.Textbox(interactive=False, lines=5, max_lines=5, autoscroll=True, show_label=False, placeholder="ここにログが表示されます...")
|
518 |
|
519 |
with gr.Column(scale=1):
|
|
|
520 |
with gr.Row():
|
521 |
+
# ▼▼▼ 変更点: 「融☆合モデルを使う」を「FNモデルモード」に変更し、最初は非表示に ▼▼▼
|
522 |
+
use_fn_model_mode_checkbox = gr.Checkbox(label="FNモデルモード", value=False, interactive=True, scale=3, visible=False)
|
523 |
+
# ▲▲▲ 変更点 ▲▲▲
|
524 |
refresh_model_list_button = gr.Button("再読込", scale=1)
|
525 |
+
selected_model_dropdown = gr.Dropdown(label="話者", choices=[], value=None, interactive=True)
|
|
|
|
|
|
|
|
|
526 |
current_styles_dropdown = gr.Dropdown(label="スタイル", choices=[], type="value", interactive=True)
|
527 |
style_weight_for_synth_slider = gr.Slider(label="スタイル強度", minimum=0.0, maximum=20.0, value=DEFAULT_STYLE_WEIGHT, step=0.1, info="初期値は推奨強度", interactive=True)
|
528 |
batch_count_slider = gr.Slider(label="生成数", value=1, minimum=1, maximum=MAX_AUDIO_OUTPUTS, step=1, interactive=True)
|
|
|
549 |
random_text_mode_slider = gr.Slider(label="分割の単位", minimum=1, maximum=4, value=1, step=1, info="1:形態素, 2:チャンク, 3:文節, 4:節", interactive=True)
|
550 |
random_text_ratio_textbox = gr.Textbox(label="カタカナ化の割合", value="0.5, 1", info="カンマ区切りで複数指定可。指定値からランダムに1つ使用。", interactive=True)
|
551 |
|
552 |
+
with gr.Tab("作業台"):
|
553 |
+
gr.Markdown("## 作業台\n読み上げタブで生成した音声をここにストックし、結合や保存ができます。最大8個まで保持できます。")
|
|
|
554 |
|
555 |
workbench_items = []
|
556 |
all_workbench_ui_components = []
|
557 |
|
558 |
with gr.Row(variant="panel"):
|
|
|
559 |
with gr.Column(scale=3):
|
560 |
with gr.Row():
|
|
|
561 |
left_workbench_col = gr.Column(scale=1)
|
|
|
562 |
right_workbench_col = gr.Column(scale=1)
|
|
|
|
|
563 |
with gr.Column(scale=1):
|
564 |
with gr.Blocks():
|
565 |
with gr.Row():
|
|
|
576 |
|
577 |
with gr.Row():
|
578 |
merge_preview_button = gr.Button("1.結合&プレビュー", variant="primary")
|
579 |
+
add_merged_to_workbench_button = gr.Button("2.作業台に追加", variant="primary")
|
580 |
delete_originals_checkbox = gr.Checkbox(label="結合時に自動で元ファイルを削除", value=False, interactive=True)
|
581 |
|
582 |
preview_audio_player = gr.Audio(label="結合結果プレビュー", interactive=False, type="filepath")
|
583 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
584 |
ITEMS_PER_COLUMN = 4
|
585 |
for i in range(MAX_WORKBENCH_ITEMS):
|
|
|
586 |
parent_column = left_workbench_col if i < ITEMS_PER_COLUMN else right_workbench_col
|
587 |
|
588 |
with parent_column:
|
|
|
597 |
"audio": audio, "info": info, "delete_btn": delete_btn
|
598 |
})
|
599 |
|
|
|
600 |
for item in workbench_items:
|
601 |
all_workbench_ui_components.extend([
|
602 |
item["row"], item["item_num_display"], item["audio"], item["info"]
|
603 |
])
|
|
|
604 |
|
605 |
|
606 |
# --- UIイベントハンドラ関数 ---
|
|
|
|
|
|
|
|
|
|
|
607 |
def load_styles_for_ui(selected_model_name: Optional[str]):
|
608 |
if not selected_model_name: return gr.update(choices=[], value=None), gr.update(value=DEFAULT_STYLE_WEIGHT), {}
|
609 |
model_path = assets_root_path / selected_model_name
|
|
|
619 |
default_weight = styles_map[first_key].get("weight", DEFAULT_STYLE_WEIGHT)
|
620 |
return gr.update(choices=display_names, value=default_display_name), gr.update(value=default_weight), styles_map
|
621 |
|
622 |
+
# ▼▼▼ 変更点: FNモデルモードに対応したモデルリスト更新関数 ▼▼▼
|
623 |
+
def action_refresh_model_list(use_fn_model_mode: bool):
|
624 |
"""モデルリストを再読み込みし、UIとバックエンドの状態を同期させる。"""
|
625 |
+
# 既存のシンボリックリンクをクリア
|
626 |
+
if assets_root_path.exists():
|
627 |
+
for item in assets_root_path.iterdir():
|
628 |
+
if item.is_symlink():
|
629 |
+
try:
|
630 |
+
item.unlink()
|
631 |
+
except OSError as e:
|
632 |
+
print(f"Failed to remove symlink {item}: {e}")
|
633 |
+
|
634 |
+
# FNモデルキャッシュパスからFNモデルを探し、シンボリックリンクを作成
|
635 |
+
fn_models_found = []
|
636 |
+
if FN_MODEL_CACHE_PATH.exists() and FN_MODEL_CACHE_PATH.is_dir():
|
637 |
+
for item in FN_MODEL_CACHE_PATH.iterdir():
|
638 |
+
if item.is_dir() and re.fullmatch(r'FN\d+', item.name, re.IGNORECASE):
|
639 |
+
fn_models_found.append(item.name)
|
640 |
+
target_link = assets_root_path / item.name
|
641 |
+
if not target_link.exists():
|
642 |
+
try:
|
643 |
+
os.symlink(item, target_link)
|
644 |
+
except OSError as e:
|
645 |
+
print(f"Warning: Could not create symlink for {item.name}: {e}")
|
646 |
+
|
647 |
+
fn_mode_checkbox_update = gr.update(visible=bool(fn_models_found))
|
648 |
+
|
649 |
+
# バックエンドのモデルリストを更新
|
650 |
model_holder.refresh()
|
651 |
|
652 |
+
model_dropdown_update = gr.update(choices=[], value=None)
|
653 |
+
if use_fn_model_mode and fn_models_found:
|
654 |
+
# FNモデルモードの場合: FNモデルを数値でソートして表示
|
655 |
+
sorted_fn_models = sorted(fn_models_found, key=get_fn_model_sort_key)
|
656 |
+
value = sorted_fn_models[0] if sorted_fn_models else None
|
657 |
+
model_dropdown_update = gr.update(choices=sorted_fn_models, value=value)
|
658 |
+
else:
|
659 |
+
# 通常モードの場合: 通常モデルとマージモデルを表示
|
660 |
+
ui_model_list = [p.name for p in assets_root_path.iterdir() if p.is_dir() and not p.is_symlink()]
|
661 |
+
formatted_choices = format_and_sort_model_names(ui_model_list)
|
662 |
+
value = formatted_choices[0][1] if formatted_choices else None
|
663 |
+
model_dropdown_update = gr.update(choices=formatted_choices, value=value)
|
664 |
+
|
665 |
+
# 選択されたモデルに基づいてスタイルUIを更新
|
666 |
+
selected_model_for_style = model_dropdown_update.value
|
667 |
+
style_dropdown_update, style_weight_update, styles_data_state_update = load_styles_for_ui(selected_model_for_style)
|
668 |
|
669 |
+
return fn_mode_checkbox_update, model_dropdown_update, style_dropdown_update, style_weight_update, styles_data_state_update
|
670 |
# ▲▲▲ 変更点 ▲▲▲
|
671 |
|
672 |
+
|
673 |
def on_model_select_change(selected_model_name: Optional[str]):
|
|
|
674 |
style_dropdown_update, style_weight_update, styles_data_state_update = load_styles_for_ui(selected_model_name)
|
675 |
return style_dropdown_update, style_weight_update, styles_data_state_update
|
|
|
676 |
|
677 |
def on_style_dropdown_select(selected_display_name: Optional[str], styles_data: Dict[str, Any]):
|
678 |
if not selected_display_name or not styles_data: return gr.update(value=DEFAULT_STYLE_WEIGHT)
|
|
|
683 |
|
684 |
def action_run_synthesis(
|
685 |
model_name: Optional[str],
|
|
|
|
|
|
|
686 |
style_display_name: Optional[str], style_weight_for_synth: float,
|
687 |
text: str, generation_mode: str, batch_count: int,
|
688 |
lang: str, seed: int, speaker: str, ref_audio: Optional[str],
|
|
|
729 |
|
730 |
all_logs = []
|
731 |
|
|
|
732 |
model_path = assets_root_path / model_name
|
733 |
files = find_safetensors_files_webui(str(model_path))
|
734 |
if not files:
|
735 |
error_outputs[0] = f"❌ [エラー] モデルフォルダ '{model_name}' に .safetensors ファイルが見つかりません。"
|
736 |
return tuple(error_outputs)
|
737 |
|
|
|
738 |
actual_model_file_to_load = str(model_path / files[0])
|
739 |
all_logs.append(f"[自動選択] 使用モデルファイル: {files[0]}")
|
|
|
740 |
|
741 |
batch_count = int(batch_count)
|
742 |
if batch_count <= 0: batch_count = 1
|
|
|
854 |
|
855 |
return tuple(final_outputs)
|
856 |
|
857 |
+
# --- 作業台イベントハンドラ ---
|
858 |
def add_to_workbench(
|
859 |
current_status: str,
|
860 |
current_workbench_list: List[Dict],
|
|
|
864 |
safe_workbench_list = current_workbench_list or []
|
865 |
|
866 |
if not audio_path or not Path(audio_path).exists():
|
867 |
+
log_messages.append("⚠️ [作業台追加エラー] 追加する音声ファイルが見つかりません。")
|
868 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
869 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
870 |
|
871 |
if any(item['audio_path'] == audio_path for item in safe_workbench_list):
|
872 |
+
log_messages.append("ℹ️ この音声はすでに作業台に存在します。")
|
873 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
874 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
875 |
|
|
|
898 |
path_to_delete.unlink()
|
899 |
except Exception as e:
|
900 |
print(f"Warning: Failed to delete old workbench audio file: {e}")
|
901 |
+
log_messages.append(f"ℹ️ 作業台のアイテムが最大数({MAX_WORKBENCH_ITEMS})に達したため、一番古いアイテムを削除しました。")
|
902 |
|
903 |
ui_updates = update_workbench_ui(updated_list)
|
904 |
+
log_messages.append("✅ 作業台に音声を追加しました。")
|
905 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
906 |
return (final_status, updated_list) + ui_updates
|
907 |
|
|
|
916 |
path_to_delete = Path(item_to_remove['audio_path'])
|
917 |
if path_to_delete.exists() and str(path_to_delete.parent) == tempfile.gettempdir():
|
918 |
path_to_delete.unlink()
|
919 |
+
log_messages.append(f"✅ 作業台からアイテム #{index_to_remove + 1} を削除し、一時ファイルをクリーンアップしました。")
|
920 |
elif path_to_delete.exists():
|
921 |
+
log_messages.append(f"✅ 作業台からアイテム #{index_to_remove + 1} を削除しました。(ファイルは保持: {path_to_delete.name})")
|
922 |
else:
|
923 |
+
log_messages.append(f"✅ 作業台からアイテム #{index_to_remove + 1} を削除しました。(関連ファイルなし)")
|
924 |
except Exception as e:
|
925 |
+
log_messages.append(f"⚠️ 作業台のアイテム #{index_to_remove + 1} のファイル削除中にエラー: {e}")
|
926 |
|
927 |
updated_list = [item for i, item in enumerate(safe_workbench_list) if i != index_to_remove]
|
928 |
ui_updates = update_workbench_ui(updated_list)
|
|
|
937 |
):
|
938 |
log_messages = []
|
939 |
if not workbench_list:
|
940 |
+
log_messages.append("⚠️ [結合プレビュー警告] 作業台に音声がありません。")
|
941 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
942 |
return final_status, None, {}
|
943 |
|
|
|
965 |
pause_duration = int(pause_ms)
|
966 |
|
967 |
if pause_duration >= 0:
|
|
|
968 |
pause_segment = AudioSegment.silent(duration=pause_duration)
|
969 |
combined_audio = segment1 + pause_segment + segment2
|
970 |
log_messages.append(f"✅ 音声 #{first_audio_num} と #{second_audio_num} を {pause_duration}ms のポーズを挟んで結合しました。")
|
971 |
else:
|
|
|
972 |
overlap_duration = abs(pause_duration)
|
|
|
|
|
973 |
max_possible_overlap = min(len(segment1), len(segment2))
|
974 |
if overlap_duration > max_possible_overlap:
|
975 |
log_messages.append(f"ℹ️ オーバーラップ長({overlap_duration}ms)が可能な最大値({max_possible_overlap}ms)を超えるため、自動的に調整されました。")
|
976 |
overlap_duration = max_possible_overlap
|
977 |
|
|
|
978 |
final_duration = len(segment1) + len(segment2) - overlap_duration
|
|
|
|
|
979 |
combined_audio = AudioSegment.silent(duration=final_duration)
|
|
|
|
|
980 |
combined_audio = combined_audio.overlay(segment1, position=0)
|
|
|
|
|
981 |
overlay_position = len(segment1) - overlap_duration
|
982 |
combined_audio = combined_audio.overlay(segment2, position=overlay_position)
|
|
|
983 |
log_messages.append(f"✅ 音声 #{first_audio_num} と #{second_audio_num} を {overlap_duration}ms 重ねて(オーバーレイして)結合しました。")
|
984 |
|
985 |
progress(1, desc="結合完了")
|
|
|
1010 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
1011 |
return final_status, str(temp_path), metadata
|
1012 |
|
|
|
|
|
|
|
|
|
1013 |
def action_add_merged_to_workbench(
|
1014 |
current_status: str,
|
1015 |
preview_data: Dict,
|
|
|
1021 |
log_messages = []
|
1022 |
safe_workbench_list = current_workbench_list or []
|
1023 |
if not preview_data or "audio_path" not in preview_data:
|
1024 |
+
log_messages.append("⚠️ [作業台追加エラー] 追加する結合済み音声がありません。先にプレビューを生成してください。")
|
1025 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
1026 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
1027 |
|
1028 |
src_path = Path(preview_data["audio_path"])
|
1029 |
if not src_path.exists():
|
1030 |
+
log_messages.append("⚠️ [作業台追加エラー] 結合済み音声ファイルが見つかりません。")
|
1031 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
1032 |
return (final_status, safe_workbench_list) + update_workbench_ui(safe_workbench_list)
|
1033 |
|
|
|
1059 |
log_messages.append(f"⚠️ 元の音声ファイル削除中にエラー: {e}")
|
1060 |
|
1061 |
final_workbench_list = [new_merged_item] + remaining_list
|
1062 |
+
log_messages.append(f"✅ 結合音声を作業台に追加し、元の音声(#{idx1+1}, #{idx2+1})を削除しました。")
|
1063 |
else:
|
1064 |
final_workbench_list = [new_merged_item] + safe_workbench_list
|
1065 |
+
log_messages.append("✅ 結合済みの音声を作業台の一番上に追加しました。")
|
1066 |
|
1067 |
if len(final_workbench_list) > MAX_WORKBENCH_ITEMS:
|
1068 |
item_to_remove = final_workbench_list.pop(-1)
|
|
|
1072 |
path_to_delete.unlink()
|
1073 |
except Exception as e:
|
1074 |
print(f"Warning: Failed to delete old workbench audio file: {e}")
|
1075 |
+
log_messages.append(f"ℹ️ 作業台が最大数({MAX_WORKBENCH_ITEMS})に達したため一番古いアイテムを削除しました。")
|
1076 |
|
1077 |
ui_updates = update_workbench_ui(final_workbench_list)
|
1078 |
final_status = (current_status + "\n" + "\n".join(log_messages)).strip()
|
|
|
1080 |
|
1081 |
|
1082 |
# --- イベントリスナー接続 ---
|
1083 |
+
# ▼▼▼ 変更点: FNモデルモードに対応したイベントリスナーの定義 ▼▼▼
|
1084 |
+
refresh_triggers = [refresh_model_list_button.click, use_fn_model_mode_checkbox.change]
|
1085 |
+
outputs_for_refresh = [use_fn_model_mode_checkbox, selected_model_dropdown, current_styles_dropdown, style_weight_for_synth_slider, all_styles_data_state]
|
1086 |
+
|
1087 |
+
for trigger in refresh_triggers:
|
1088 |
+
trigger(action_refresh_model_list,
|
1089 |
+
inputs=[use_fn_model_mode_checkbox],
|
1090 |
+
outputs=outputs_for_refresh)
|
1091 |
|
1092 |
selected_model_dropdown.change(on_model_select_change,
|
1093 |
inputs=[selected_model_dropdown],
|
|
|
1105 |
generate_button.click(
|
1106 |
fn=action_run_synthesis,
|
1107 |
inputs=[
|
1108 |
+
selected_model_dropdown,
|
|
|
1109 |
current_styles_dropdown, style_weight_for_synth_slider,
|
1110 |
text_input, generation_mode_radio, batch_count_slider,
|
1111 |
language_dropdown, seed_input, speaker_name_textbox,
|
|
|
1120 |
)
|
1121 |
|
1122 |
for i in range(MAX_AUDIO_OUTPUTS):
|
|
|
|
|
|
|
1123 |
to_workbench_buttons[i].click(
|
1124 |
fn=add_to_workbench,
|
1125 |
inputs=[
|
|
|
1148 |
outputs=[status_textbox, preview_audio_player, merged_preview_state]
|
1149 |
)
|
1150 |
|
|
|
|
|
|
|
|
|
1151 |
add_merged_to_workbench_button.click(
|
1152 |
fn=action_add_merged_to_workbench,
|
1153 |
inputs=[
|
|
|
1163 |
|
1164 |
player_width_slider.release(lambda w: f"<script>document.documentElement.style.setProperty('--audio-width', '{w}px');</script>", inputs=[player_width_slider], outputs=[js_injector_html])
|
1165 |
|
1166 |
+
app.load(action_refresh_model_list,
|
1167 |
+
inputs=[use_fn_model_mode_checkbox],
|
1168 |
+
outputs=outputs_for_refresh)
|
|
|
|
|
1169 |
# ▲▲▲ 変更点 ▲▲▲
|
1170 |
return app
|
1171 |
|
|
|
1173 |
if __name__ == "__main__":
|
1174 |
if Path("model_assets").exists(): shutil.rmtree("model_assets")
|
1175 |
|
1176 |
+
# ▼▼▼ 変更点: FNモデルキャッシュパスの作成とGradioへの許可 ▼▼▲
|
1177 |
+
# モックFNモデルを作成してテスト
|
1178 |
+
FN_MODEL_CACHE_PATH.mkdir(exist_ok=True, parents=True)
|
1179 |
+
(FN_MODEL_CACHE_PATH / "FN1").mkdir(exist_ok=True)
|
1180 |
+
(FN_MODEL_CACHE_PATH / "FN10").mkdir(exist_ok=True)
|
1181 |
+
(FN_MODEL_CACHE_PATH / "FN2").mkdir(exist_ok=True)
|
1182 |
+
|
1183 |
mock_model_holder = TTSModelHolder()
|
1184 |
print(f"Initial models loaded by TTSModelHolder: {mock_model_holder.model_names}")
|
1185 |
app = create_synthesis_app(mock_model_holder)
|
|
|
1188 |
assets_dir_path.mkdir(exist_ok=True)
|
1189 |
allowed_paths = [str(assets_dir_path)]
|
1190 |
|
1191 |
+
# FNモデルのキャッシュパスへのアクセスを許可
|
1192 |
+
if FN_MODEL_CACHE_PATH.exists() and FN_MODEL_CACHE_PATH.is_dir():
|
1193 |
+
allowed_paths.append(str(FN_MODEL_CACHE_PATH.resolve()))
|
1194 |
+
else:
|
1195 |
+
try:
|
1196 |
+
FN_MODEL_CACHE_PATH.mkdir(exist_ok=True, parents=True)
|
1197 |
+
allowed_paths.append(str(FN_MODEL_CACHE_PATH.resolve()))
|
1198 |
+
print(f"Created FN model cache directory at: {FN_MODEL_CACHE_PATH}")
|
1199 |
+
except OSError as e:
|
1200 |
+
print(f"Could not create FN model cache directory at {FN_MODEL_CACHE_PATH}: {e}")
|
1201 |
|
1202 |
output_dir_path = Path("output").resolve()
|
1203 |
(output_dir_path / "normal").mkdir(exist_ok=True, parents=True)
|
|
|
1209 |
|
1210 |
print(f"Gradioに次のパスへのアクセスを許可します: {', '.join(allowed_paths)}")
|
1211 |
|
1212 |
+
app.launch(allowed_paths=allowed_paths)
|
1213 |
+
# ▲▲▲ 変更点 ▲▲▲
|