Update app.py

app.py

@@ -1,344 +1,365 @@
-import gradio as gr
-import subprocess
+import asyncio
+import datetime
+import logging
 import os
-import sys
-import yaml
-from pathlib import Path
 import time
-import threading
-import tempfile
-import shutil
-import gc # ガベージコレクション用
-
-# --- 定数 ---
-# Dockerfile内でクローンされるパスに合わせる
-SCRIPT_DIR = Path(__file__).parent
-SBV2_REPO_PATH = SCRIPT_DIR / "Style-Bert-VITS2"
-# ダウンロード用ファイルの一時置き場 (コンテナ内に作成)
-OUTPUT_DIR = SCRIPT_DIR / "outputs"
-
-# --- ヘルパー関数 ---
-def add_sbv2_to_path():
-    """Style-Bert-VITS2リポジトリのパスを sys.path に追加"""
-    repo_path_str = str(SBV2_REPO_PATH.resolve())
-    if SBV2_REPO_PATH.exists() and repo_path_str not in sys.path:
-        sys.path.insert(0, repo_path_str)
-        print(f"Added {repo_path_str} to sys.path")
-    elif not SBV2_REPO_PATH.exists():
-        print(f"Warning: Style-Bert-VITS2 repository not found at {SBV2_REPO_PATH}")
-
-def stream_process_output(process, log_list):
-    """サブプロセスの標準出力/エラーをリアルタイムでリストに追加"""
-    try:
-        if process.stdout:
-            for line in iter(process.stdout.readline, ''):
-                log_list.append(line.strip()) # 余分な改行を削除
-        if process.stderr:
-             for line in iter(process.stderr.readline, ''):
-                processed_line = f"stderr: {line.strip()}"
-                # 警告はそのまま、他はエラーとして強調 (任意)
-                if "warning" not in line.lower():
-                     processed_line = f"ERROR (stderr): {line.strip()}"
-                log_list.append(processed_line)
-    except Exception as e:
-        log_list.append(f"Error reading process stream: {e}")
-
-# --- Gradio アプリのバックエンド関数 ---
-def convert_safetensors_to_onnx_gradio(
-    safetensors_file_obj,
-    config_file_obj,
-    style_vectors_file_obj
-    ): # gr.Progress は削除
-    """
-    アップロードされたSafetensors, config.json, style_vectors.npy を使って
-    ONNXに変換し、結果をダウンロード可能にする。
-    """
-    log = ["Starting ONNX conversion..."]
-    # 初期状態ではダウンロードファイルは空
-    yield "\n".join(log), None
-
-    # --- ファイルアップロードの検証 ---
-    if safetensors_file_obj is None:
-        log.append("❌ Error: Safetensors file is missing. Please upload the .safetensors file.")
-        yield "\n".join(log), None
-        return
-    if config_file_obj is None:
-        log.append("❌ Error: config.json file is missing. Please upload the config.json file.")
-        yield "\n".join(log), None
-        return
-    if style_vectors_file_obj is None:
-        log.append("❌ Error: style_vectors.npy file is missing. Please upload the style_vectors.npy file.")
-        yield "\n".join(log), None
-        return
-
-    # --- Style-Bert-VITS2 パスの確認 ---
-    add_sbv2_to_path()
-    if not SBV2_REPO_PATH.exists():
-        log.append(f"❌ Error: Style-Bert-VITS2 repository not found at {SBV2_REPO_PATH}. Check Space build logs.")
-        yield "\n".join(log), None
-        return
-
-    # --- 出力ディレクトリ作成 ---
-    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
-
-    onnx_output_path_str = None # 最終的なONNXファイルパス (文字列)
-    current_log = log[:] # ログリストをコピー
-
-    try:
-        # --- 一時ディレクトリを作成して処理 ---
-        with tempfile.TemporaryDirectory() as temp_dir:
-            temp_dir_path = Path(temp_dir)
-            current_log.append(f"📁 Created temporary directory: {temp_dir_path}")
-            yield "\n".join(current_log), None # UI更新
-
-            # --- SBV2が期待するディレクトリ構造を一時ディレクトリ内に作成 ---
-            # モデル名を .safetensors ファイル名から取得 (拡張子なし)
-            safetensors_filename = Path(safetensors_file_obj.name).name
-            if not safetensors_filename.lower().endswith(".safetensors"):
-                 current_log.append(f"❌ Error: Invalid safetensors filename: {safetensors_filename}")
-                 yield "\n".join(current_log), None
-                 return
-            model_name = Path(safetensors_filename).stem # 拡張子を除いた部分
-
-            # assets_root を一時ディレクトリ自体にする
-            assets_root = temp_dir_path
-            # assets_root の下に model_name のディレクトリを作成
-            model_dir_in_temp = assets_root / model_name
-            model_dir_in_temp.mkdir(exist_ok=True)
-            current_log.append(f"   - Created model directory: {model_dir_in_temp.relative_to(assets_root)}")
-            yield "\n".join(current_log), None
-
-            # --- 3つのファイルを model_dir_in_temp にコピー ---
-            files_to_copy = {
-                "safetensors": safetensors_file_obj,
-                "config.json": config_file_obj,
-                "style_vectors.npy": style_vectors_file_obj,
-            }
-            copied_paths = {}
-
-            for file_key, file_obj in files_to_copy.items():
-                original_filename = Path(file_obj.name).name
-                # ファイル名の基本的な検証 (サニタイズはより厳密に行うことも可能)
-                if "/" in original_filename or "\\" in original_filename or ".." in original_filename:
-                     current_log.append(f"❌ Error: Invalid characters found in filename: {original_filename}")
-                     yield "\n".join(current_log), None
-                     return # tryブロックを抜ける
-                # 期待されるファイル名と一致しているか確認 (config と style_vectors)
-                if file_key == "config.json" and original_filename.lower() != "config.json":
-                    current_log.append(f"⚠️ Warning: Uploaded JSON file name is '{original_filename}', expected 'config.json'. Using uploaded name.")
-                if file_key == "style_vectors.npy" and original_filename.lower() != "style_vectors.npy":
-                    current_log.append(f"⚠️ Warning: Uploaded NPY file name is '{original_filename}', expected 'style_vectors.npy'. Using uploaded name.")
-
-                destination_path = model_dir_in_temp / original_filename
-                try:
-                    shutil.copy(file_obj.name, destination_path)
-                    current_log.append(f"   - Copied '{original_filename}' to model directory.")
-                    # .safetensorsファイルのパスを保存しておく
-                    if file_key == "safetensors":
-                        copied_paths["safetensors"] = destination_path
-                except Exception as e:
-                     current_log.append(f"❌ Error copying file '{original_filename}': {e}")
-                     yield "\n".join(current_log), None
-                     return # tryブロックを抜ける
-                yield "\n".join(current_log), None # 各ファイルコピー後にUI更新
-
-            # safetensorsファイルがコピーされたか確認
-            temp_safetensors_path = copied_paths.get("safetensors")
-            if not temp_safetensors_path:
-                current_log.append("❌ Error: Failed to locate the copied safetensors file in the temporary directory.")
-                yield "\n".join(current_log), None
-                return
-
-            current_log.append(f"✅ All required files copied to temporary model directory.")
-            current_log.append(f"   - Using temporary assets_root: {assets_root}")
-            yield "\n".join(current_log), None
-
-            # --- paths.yml を一時的に設定 ---
-            config_path = SBV2_REPO_PATH / "configs" / "paths.yml"
-            config_path.parent.mkdir(parents=True, exist_ok=True)
-            # dataset_root は今回は使わないが設定はしておく (assets_rootと同じ場所)
-            paths_config = {"dataset_root": str(assets_root.resolve()), "assets_root": str(assets_root.resolve())}
-            with open(config_path, "w", encoding="utf-8") as f:
-                yaml.dump(paths_config, f)
-            current_log.append(f"   - Saved temporary paths config to {config_path}")
-            yield "\n".join(current_log), None
-
-            # --- ONNX変換スクリプト実行 ---
-            current_log.append(f"\n🚀 Starting ONNX conversion script for model '{model_name}'...")
-            convert_script = SBV2_REPO_PATH / "convert_onnx.py"
-            if not convert_script.exists():
-                 current_log.append(f"❌ Error: convert_onnx.py not found at '{convert_script}'. Check repository setup.")
-                 yield "\n".join(current_log), None
-                 return # tryブロックを抜ける
-
-            python_executable = sys.executable
-            command = [
-                python_executable,
-                str(convert_script.resolve()),
-                "--model",
-                str(temp_safetensors_path.resolve()) # 一時ディレクトリ内の .safetensors パス
-            ]
-            current_log.append(f"\n   Running command: {' '.join(command)}")
-            yield "\n".join(current_log), None
-
-            process_env = os.environ.copy()
-            process = subprocess.Popen(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
-                                       text=True, encoding='utf-8', errors='replace',
-                                       cwd=SBV2_REPO_PATH, # スクリプトの場所で実行
-                                       env=process_env)
-
-            # ログ出力用リスト (スレッドと共有)
-            process_output_lines = ["\n--- Conversion Script Output ---"]
-            thread = threading.Thread(target=stream_process_output, args=(process, process_output_lines))
-            thread.start()
+import traceback
 
-            # 進捗表示のためのループ
-            while thread.is_alive():
-                yield "\n".join(current_log + process_output_lines), None
-                time.sleep(0.3) # 更新頻度
-
-            # スレッド終了待ちとプロセス終了待ち
-            thread.join()
-            try:
-                 process.wait(timeout=12000) # 20分タイムアウト (モデルサイズにより調整)
-            except subprocess.TimeoutExpired:
-                 current_log.extend(process_output_lines) # ここまでのログを追加
-                 current_log.append("\n❌ Error: Conversion process timed out after 20 minutes.")
-                 process.kill()
-                 yield "\n".join(current_log), None
-                 return # tryブロックを抜ける
-
-            # 最終的なプロセス出力を取得
-            final_stdout, final_stderr = process.communicate()
-            if final_stdout:
-                 process_output_lines.extend(final_stdout.strip().split('\n'))
-            if final_stderr:
-                 processed_stderr = []
-                 for line in final_stderr.strip().split('\n'):
-                     processed_line = f"stderr: {line.strip()}"
-                     if "warning" not in line.lower() and line.strip(): # 空行と警告以外
-                         processed_line = f"ERROR (stderr): {line.strip()}"
-                     processed_stderr.append(processed_line)
-                 if any(line.startswith("ERROR") for line in processed_stderr):
-                      process_output_lines.append("--- Errors/Warnings (stderr) ---")
-                      process_output_lines.extend(processed_stderr)
-                      process_output_lines.append("-----------------------------")
-                 elif processed_stderr: # 警告のみの場合
-                      process_output_lines.append("--- Warnings (stderr) ---")
-                      process_output_lines.extend(processed_stderr)
-                      process_output_lines.append("------------------------")
-
-
-            # 全てのプロセスログをメインログに追加
-            current_log.extend(process_output_lines)
-            current_log.append("--- End Script Output ---")
-            current_log.append("\n-------------------------------")
-
-            # --- 結果の確認と出力ファイルのコピー ---
-            if process.returncode == 0:
-                current_log.append("✅ ONNX conversion command finished successfully.")
-                # 期待されるONNXファイルパス (入力と同じディレクトリ内)
-                expected_onnx_path_in_temp = temp_safetensors_path.with_suffix(".onnx")
-
-                if expected_onnx_path_in_temp.exists():
-                    current_log.append(f"   - Found converted ONNX file: {expected_onnx_path_in_temp.name}")
-                    # 一時ディレクトリから永続的な出力ディレクトリにコピー
-                    final_onnx_path = OUTPUT_DIR / expected_onnx_path_in_temp.name
-                    try:
-                        shutil.copy(expected_onnx_path_in_temp, final_onnx_path)
-                        current_log.append(f"   - Copied ONNX file for download to: {final_onnx_path.relative_to(SCRIPT_DIR)}")
-                        onnx_output_path_str = str(final_onnx_path) # ダウンロード用ファイルパスを設定
-                    except Exception as e:
-                        current_log.append(f"❌ Error copying ONNX file to output directory: {e}")
-                else:
-                    current_log.append(f"⚠️ Warning: Expected ONNX file not found at '{expected_onnx_path_in_temp.name}'. Check script output above.")
-            else:
-                current_log.append(f"❌ ONNX conversion command failed with return code {process.returncode}.")
-                current_log.append("   Please check the logs above for errors (especially lines starting with 'ERROR').")
-
-            # 一時ディレクトリが自動で削除される前に最終結果をyield
-            yield "\n".join(current_log), onnx_output_path_str
-
-    except FileNotFoundError as e:
-        current_log.append(f"\n❌ Error: A required command or file was not found: {e.filename}. Check Dockerfile setup and PATH.")
-        current_log.append(f"{e}")
-        yield "\n".join(current_log), None
-    except Exception as e:
-        current_log.append(f"\n❌ An unexpected error occurred: {e}")
-        import traceback
-        current_log.append(traceback.format_exc())
-        yield "\n".join(current_log), None
-    finally:
-        # ガベージコレクション
-        gc.collect()
-        print("Conversion function finished.") # サーバーログ用
-        # 最後のyieldでUIを最終状態に更新
-        # yield "\n".join(current_log), onnx_output_path_str # tryブロック内で既に返している
-
-
-# --- Gradio Interface ---
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# Style-Bert-VITS2 Safetensors to ONNX Converter")
-    gr.Markdown(
-        "Upload your model's `.safetensors`, `config.json`, and `style_vectors.npy` files. "
-        "The application will convert the model to ONNX format, and you can download the resulting `.onnx` file."
-    )
-    gr.Markdown(
-        "_(Environment setup is handled automatically when this Space starts.)_"
+import edge_tts
+import gradio as gr
+import librosa
+import torch
+from fairseq import checkpoint_utils
+
+from config import Config
+from lib.infer_pack.models import (
+    SynthesizerTrnMs256NSFsid,
+    SynthesizerTrnMs256NSFsid_nono,
+    SynthesizerTrnMs768NSFsid,
+    SynthesizerTrnMs768NSFsid_nono,
+)
+from rmvpe import RMVPE
+from vc_infer_pipeline import VC
+
+logging.getLogger("fairseq").setLevel(logging.WARNING)
+logging.getLogger("numba").setLevel(logging.WARNING)
+logging.getLogger("markdown_it").setLevel(logging.WARNING)
+logging.getLogger("urllib3").setLevel(logging.WARNING)
+logging.getLogger("matplotlib").setLevel(logging.WARNING)
+
+limitation = os.getenv("SYSTEM") == "spaces"
+
+config = Config()
+
+edge_output_filename = "edge_output.mp3"
+tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
+tts_voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]
+
+model_root = "weights"
+models = [
+    d for d in os.listdir(model_root) if os.path.isdir(os.path.join(model_root, d))
+]
+if len(models) == 0:
+    raise ValueError("No model found in `weights` folder")
+models.sort()
+
+
+def model_data(model_name):
+    # global n_spk, tgt_sr, net_g, vc, cpt, version, index_file
+    pth_files = [
+        os.path.join(model_root, model_name, f)
+        for f in os.listdir(os.path.join(model_root, model_name))
+        if f.endswith(".pth")
+    ]
+    if len(pth_files) == 0:
+        raise ValueError(f"No pth file found in {model_root}/{model_name}")
+    pth_path = pth_files[0]
+    print(f"Loading {pth_path}")
+    cpt = torch.load(pth_path, map_location="cpu")
+    tgt_sr = cpt["config"][-1]
+    cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]  # n_spk
+    if_f0 = cpt.get("f0", 1)
+    version = cpt.get("version", "v1")
+    if version == "v1":
+        if if_f0 == 1:
+            net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half)
+        else:
+            net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
+    elif version == "v2":
+        if if_f0 == 1:
+            net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half)
+        else:
+            net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
+    else:
+        raise ValueError("Unknown version")
+    del net_g.enc_q
+    net_g.load_state_dict(cpt["weight"], strict=False)
+    print("Model loaded")
+    net_g.eval().to(config.device)
+    if config.is_half:
+        net_g = net_g.half()
+    else:
+        net_g = net_g.float()
+    vc = VC(tgt_sr, config)
+    # n_spk = cpt["config"][-3]
+
+    index_files = [
+        os.path.join(model_root, model_name, f)
+        for f in os.listdir(os.path.join(model_root, model_name))
+        if f.endswith(".index")
+    ]
+    if len(index_files) == 0:
+        print("No index file found")
+        index_file = ""
+    else:
+        index_file = index_files[0]
+        print(f"Index file found: {index_file}")
+
+    return tgt_sr, net_g, vc, version, index_file, if_f0
+
+
+def load_hubert():
+    global hubert_model
+    models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
+        ["hubert_base.pt"],
+        suffix="",
     )
+    hubert_model = models[0]
+    hubert_model = hubert_model.to(config.device)
+    if config.is_half:
+        hubert_model = hubert_model.half()
+    else:
+        hubert_model = hubert_model.float()
+    return hubert_model.eval()
+
+
+print("Loading hubert model...")
+hubert_model = load_hubert()
+print("Hubert model loaded.")
+
+print("Loading rmvpe model...")
+rmvpe_model = RMVPE("rmvpe.pt", config.is_half, config.device)
+print("rmvpe model loaded.")
+
+
+def tts(
+    model_name,
+    speed,
+    tts_text,
+    tts_voice,
+    f0_up_key,
+    f0_method,
+    index_rate,
+    protect,
+    filter_radius=3,
+    resample_sr=0,
+    rms_mix_rate=0.25,
+):
+    print("------------------")
+    print(datetime.datetime.now())
+    print("tts_text:")
+    print(tts_text)
+    print(f"tts_voice: {tts_voice}")
+    print(f"Model name: {model_name}")
+    print(f"F0: {f0_method}, Key: {f0_up_key}, Index: {index_rate}, Protect: {protect}")
+    try:
+        if limitation and len(tts_text) > 280:
+            print("Error: Text too long")
+            return (
+                f"Text characters should be at most 280 in this huggingface space, but got {len(tts_text)} characters.",
+                None,
+                None,
+            )
+        tgt_sr, net_g, vc, version, index_file, if_f0 = model_data(model_name)
+        t0 = time.time()
+        if speed >= 0:
+            speed_str = f"+{speed}%"
+        else:
+            speed_str = f"{speed}%"
+        asyncio.run(
+            edge_tts.Communicate(
+                tts_text, "-".join(tts_voice.split("-")[:-1]), rate=speed_str
+            ).save(edge_output_filename)
+        )
+        t1 = time.time()
+        edge_time = t1 - t0
+        audio, sr = librosa.load(edge_output_filename, sr=16000, mono=True)
+        duration = len(audio) / sr
+        print(f"Audio duration: {duration}s")
+        if limitation and duration >= 20:
+            print("Error: Audio too long")
+            return (
+                f"Audio should be less than 20 seconds in this huggingface space, but got {duration}s.",
+                edge_output_filename,
+                None,
+            )
 
+        f0_up_key = int(f0_up_key)
+
+        if not hubert_model:
+            load_hubert()
+        if f0_method == "rmvpe":
+            vc.model_rmvpe = rmvpe_model
+        times = [0, 0, 0]
+        audio_opt = vc.pipeline(
+            hubert_model,
+            net_g,
+            0,
+            audio,
+            edge_output_filename,
+            times,
+            f0_up_key,
+            f0_method,
+            index_file,
+            # file_big_npy,
+            index_rate,
+            if_f0,
+            filter_radius,
+            tgt_sr,
+            resample_sr,
+            rms_mix_rate,
+            version,
+            protect,
+            None,
+        )
+        if tgt_sr != resample_sr >= 16000:
+            tgt_sr = resample_sr
+        info = f"Success. Time: edge-tts: {edge_time}s, npy: {times[0]}s, f0: {times[1]}s, infer: {times[2]}s"
+        print(info)
+        return (
+            info,
+            edge_output_filename,
+            (tgt_sr, audio_opt),
+        )
+    except EOFError:
+        info = (
+            "It seems that the edge-tts output is not valid. "
+            "This may occur when the input text and the speaker do not match. "
+            "For example, maybe you entered Japanese (without alphabets) text but chose non-Japanese speaker?"
+        )
+        print(info)
+        return info, None, None
+    except:
+        info = traceback.format_exc()
+        print(info)
+        return info, None, None
+
+
+initial_md = """
+# RVC text-to-speech webui
+
+This is a text-to-speech webui of RVC models.
+
+Input text ➡[(edge-tts)](https://github.com/rany2/edge-tts)➡ Speech mp3 file ➡[(RVC)](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)➡ Final output
+"""
+
+app = gr.Blocks()
+with app:
+    gr.Markdown(initial_md)
     with gr.Row():
-        with gr.Column(scale=1):
-            gr.Markdown("### 1. Upload Model Files")
-            safetensors_upload = gr.File(
-                label="Safetensors Model (.safetensors)",
-                file_types=[".safetensors"],
+        with gr.Column():
+            model_name = gr.Dropdown(label="Model", choices=models, value=models[0])
+            f0_key_up = gr.Number(
+                label="Transpose (the best value depends on the models and speakers)",
+                value=0,
             )
-            config_upload = gr.File(
-                label="Config File (config.json)",
-                file_types=[".json"],
+        with gr.Column():
+            f0_method = gr.Radio(
+                label="Pitch extraction method (pm: very fast, low quality, rmvpe: a little slow, high quality)",
+                choices=["pm", "rmvpe"],  # harvest and crepe is too slow
+                value="rmvpe",
+                interactive=True,
             )
-            style_vectors_upload = gr.File(
-                label="Style Vectors (style_vectors.npy)",
-                file_types=[".npy"],
+            index_rate = gr.Slider(
+                minimum=0,
+                maximum=1,
+                label="Index rate",
+                value=1,
+                interactive=True,
             )
-
-            convert_button = gr.Button("2. Convert to ONNX", variant="primary", elem_id="convert_button")
-
-            gr.Markdown("---")
-            gr.Markdown("### 3. Download Result")
-            onnx_download = gr.File(
-                label="ONNX Model (.onnx)",
-                interactive=False, # 出力専用
+            protect0 = gr.Slider(
+                minimum=0,
+                maximum=0.5,
+                label="Protect",
+                value=0.33,
+                step=0.01,
+                interactive=True,
             )
-            gr.Markdown(
-                "**Note:** Conversion can take **several minutes** (5-20+ min depending on model size and hardware). "
-                "Please be patient. The log on the right shows the progress."
+    with gr.Row():
+        with gr.Column():
+            tts_voice = gr.Dropdown(
+                label="Edge-tts speaker (format: language-Country-Name-Gender)",
+                choices=tts_voices,
+                allow_custom_value=False,
+                value="ja-JP-NanamiNeural-Female",
             )
-
-        with gr.Column(scale=2):
-            output_log = gr.Textbox(
-                label="Conversion Log",
-                lines=30, # 高さをさらに増やす
-                interactive=False,
-                autoscroll=True,
-                max_lines=2000 # ログが増える可能性
+            speed = gr.Slider(
+                minimum=-100,
+                maximum=100,
+                label="Speech speed (%)",
+                value=0,
+                step=10,
+                interactive=True,
             )
-
-    # ボタンクリック時のアクション設定
-    convert_button.click(
-        convert_safetensors_to_onnx_gradio,
-        inputs=[safetensors_upload, config_upload, style_vectors_upload],
-        outputs=[output_log, onnx_download] # ログとダウンロードファイルの2つを出力
-    )
-
-# --- アプリの起動 ---
-if __name__ == "__main__":
-    # Style-Bert-VITS2 へのパスを追加
-    add_sbv2_to_path()
-    # 出力ディレクトリ作成 (存在確認含む)
-    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
-    print(f"Output directory: {OUTPUT_DIR.resolve()}")
-
-    # Gradioアプリを起動
-    demo.queue().launch()
+            tts_text = gr.Textbox(label="Input Text", value="これは日本語テキストから音声への変換デモです。")
+        with gr.Column():
+            but0 = gr.Button("Convert", variant="primary")
+            info_text = gr.Textbox(label="Output info")
+        with gr.Column():
+            edge_tts_output = gr.Audio(label="Edge Voice", type="filepath")
+            tts_output = gr.Audio(label="Result")
+        but0.click(
+            tts,
+            [
+                model_name,
+                speed,
+                tts_text,
+                tts_voice,
+                f0_key_up,
+                f0_method,
+                index_rate,
+                protect0,
+            ],
+            [info_text, edge_tts_output, tts_output],
+        )
+    with gr.Row():
+        examples = gr.Examples(
+            examples_per_page=100,
+            examples=[
+                ["これは日本語テキストから音声への変換デモです。", "ja-JP-NanamiNeural-Female"],
+                [
+                    "This is an English text to speech conversation demo.",
+                    "en-US-AriaNeural-Female",
+                ],
+                ["这是一个中文文本到语音的转换演示。", "zh-CN-XiaoxiaoNeural-Female"],
+                ["한국어 텍스트에서 음성으로 변환하는 데모입니다.", "ko-KR-SunHiNeural-Female"],
+                [
+                    "Il s'agit d'une démo de conversion du texte français à la parole.",
+                    "fr-FR-DeniseNeural-Female",
+                ],
+                [
+                    "Dies ist eine Demo zur Umwandlung von Deutsch in Sprache.",
+                    "de-DE-AmalaNeural-Female",
+                ],
+                [
+                    "Tämä on suomenkielinen tekstistä puheeksi -esittely.",
+                    "fi-FI-NooraNeural-Female",
+                ],
+                [
+                    "Это демонстрационный пример преобразования русского текста в речь.",
+                    "ru-RU-SvetlanaNeural-Female",
+                ],
+                [
+                    "Αυτή είναι μια επίδειξη μετατροπής ελληνικού κειμένου σε ομιλία.",
+                    "el-GR-AthinaNeural-Female",
+                ],
+                [
+                    "Esta es una demostración de conversión de texto a voz en español.",
+                    "es-ES-ElviraNeural-Female",
+                ],
+                [
+                    "Questa è una dimostrazione di sintesi vocale in italiano.",
+                    "it-IT-ElsaNeural-Female",
+                ],
+                [
+                    "Esta é uma demonstração de conversão de texto em fala em português.",
+                    "pt-PT-RaquelNeural-Female",
+                ],
+                [
+                    "Це демонстрація тексту до мовлення українською мовою.",
+                    "uk-UA-PolinaNeural-Female",
+                ],
+                [
+                    "هذا عرض توضيحي عربي لتحويل النص إلى كلام.",
+                    "ar-EG-SalmaNeural-Female",
+                ],
+                [
+                    "இது தமிழ் உரையிலிருந்து பேச்சு மாற்ற டெமோ.",
+                    "ta-IN-PallaviNeural-Female",
+                ],
+            ],
+            inputs=[tts_text, tts_voice],
+        )
+
+
+app.launch(inbrowser=True)