Update app.py

app.py

@@ -3,6 +3,7 @@ If running this app in WSL2, you need to run the following command in the WSL2 t
 
 ip addr show eth0 | grep "inet\b" | awk '{print $2}' | cut -d/ -f1
 """
+
 import asyncio
 import logging
 import os
@@ -98,27 +99,36 @@ version = tts_config.version
 
 
 @spaces.GPU
-def inference(text, text_lang,
-              ref_audio_path,
-              aux_ref_audio_paths,
-              prompt_text,
-              prompt_lang, top_k,
-              top_p, temperature,
-              text_split_method, batch_size,
-              speed_factor, ref_text_free,
-              split_bucket, fragment_interval,
-              seed, keep_random, parallel_infer,
-              repetition_penalty
-              ):
-
+def inference(
+    text,
+    text_lang,
+    ref_audio_path,
+    aux_ref_audio_paths,
+    prompt_text,
+    prompt_lang,
+    top_k,
+    top_p,
+    temperature,
+    text_split_method,
+    batch_size,
+    speed_factor,
+    ref_text_free,
+    split_bucket,
+    fragment_interval,
+    seed,
+    keep_random,
+    parallel_infer,
+    repetition_penalty,
+):
     seed = -1 if keep_random else seed
-    actual_seed = seed if seed not in [-1,
-                                       "", None] else random.randrange(1 << 32)
+    actual_seed = seed if seed not in [-1, "", None] else random.randrange(1 << 32)
     inputs = {
         "text": text,
         "text_lang": dict_language[text_lang],
         "ref_audio_path": ref_audio_path,
-        "aux_ref_audio_paths": [item.name for item in aux_ref_audio_paths] if aux_ref_audio_paths is not None else [],
+        "aux_ref_audio_paths": [item.name for item in aux_ref_audio_paths]
+        if aux_ref_audio_paths is not None
+        else [],
         "prompt_text": prompt_text if not ref_text_free else "",
         "prompt_lang": dict_language[prompt_lang],
         "top_k": top_k,
@@ -140,22 +150,28 @@ def inference(text, text_lang,
 
 def custom_sort_key(s):
     # 使用正则表达式提取字符串中的数字部分和非数字部分
-    parts = re.split('(\d+)', s)
+    parts = re.split("(\d+)", s)
     # 将数字部分转换为整数，非数字部分保持不变
     parts = [int(part) if part.isdigit() else part for part in parts]
     return parts
 
 
 def change_choices():
-    SoVITS_names, GPT_names = get_weights_names(
-        GPT_weight_root, SoVITS_weight_root)
-    return {"choices": sorted(SoVITS_names, key=custom_sort_key), "__type__": "update"}, {"choices": sorted(GPT_names, key=custom_sort_key), "__type__": "update"}
+    SoVITS_names, GPT_names = get_weights_names(GPT_weight_root, SoVITS_weight_root)
+    return {
+        "choices": sorted(SoVITS_names, key=custom_sort_key),
+        "__type__": "update",
+    }, {"choices": sorted(GPT_names, key=custom_sort_key), "__type__": "update"}
 
 
 pretrained_sovits_name = [
-    "GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth", "GPT_SoVITS/pretrained_models/s2G488k.pth"]
-pretrained_gpt_name = ["GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt",
-                       "GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt"]
+    "GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth",
+    "GPT_SoVITS/pretrained_models/s2G488k.pth",
+]
+pretrained_gpt_name = [
+    "GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt",
+    "GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt",
+]
 _ = [[], []]
 for i in range(2):
     if os.path.exists(pretrained_gpt_name[i]):
@@ -166,7 +182,7 @@ pretrained_gpt_name, pretrained_sovits_name = _
 
 SoVITS_weight_root = ["SoVITS_weights_v2", "SoVITS_weights"]
 GPT_weight_root = ["GPT_weights_v2", "GPT_weights"]
-for path in SoVITS_weight_root+GPT_weight_root:
+for path in SoVITS_weight_root + GPT_weight_root:
     os.makedirs(path, exist_ok=True)
 
 
@@ -184,8 +200,7 @@ def get_weights_names(GPT_weight_root, SoVITS_weight_root):
     return SoVITS_names, GPT_names
 
 
-SoVITS_names, GPT_names = get_weights_names(
-    GPT_weight_root, SoVITS_weight_root)
+SoVITS_names, GPT_names = get_weights_names(GPT_weight_root, SoVITS_weight_root)
 
 
 def change_sovits_weights(sovits_path, prompt_language=None, text_language=None):
@@ -194,19 +209,29 @@ def change_sovits_weights(sovits_path, prompt_language=None, text_language=None)
     dict_language = dict_language_v2
     if prompt_language is not None and text_language is not None:
         if prompt_language in list(dict_language.keys()):
-            prompt_text_update, prompt_language_update = {'__type__': 'update'},  {
-                '__type__': 'update', 'value': prompt_language}
+            prompt_text_update, prompt_language_update = (
+                {"__type__": "update"},
+                {"__type__": "update", "value": prompt_language},
+            )
         else:
-            prompt_text_update = {'__type__': 'update', 'value': ''}
-            prompt_language_update = {
-                '__type__': 'update', 'value': i18n("中文")}
+            prompt_text_update = {"__type__": "update", "value": ""}
+            prompt_language_update = {"__type__": "update", "value": i18n("中文")}
         if text_language in list(dict_language.keys()):
-            text_update, text_language_update = {'__type__': 'update'}, {
-                '__type__': 'update', 'value': text_language}
+            text_update, text_language_update = (
+                {"__type__": "update"},
+                {"__type__": "update", "value": text_language},
+            )
         else:
-            text_update = {'__type__': 'update', 'value': ''}
-            text_language_update = {'__type__': 'update', 'value': i18n("中文")}
-        return {'__type__': 'update', 'choices': list(dict_language.keys())}, {'__type__': 'update', 'choices': list(dict_language.keys())}, prompt_text_update, prompt_language_update, text_update, text_language_update
+            text_update = {"__type__": "update", "value": ""}
+            text_language_update = {"__type__": "update", "value": i18n("中文")}
+        return (
+            {"__type__": "update", "choices": list(dict_language.keys())},
+            {"__type__": "update", "choices": list(dict_language.keys())},
+            prompt_text_update,
+            prompt_language_update,
+            text_update,
+            text_language_update,
+        )
 
 
 async def create_app():
@@ -221,104 +246,215 @@ async def create_app():
             # with gr.Group():
             gr.Markdown(value=i18n("模型切换"))
             with gr.Row():
-                GPT_dropdown = gr.Dropdown(label=i18n("GPT模型列表"), choices=sorted(
-                    GPT_names, key=custom_sort_key), value=gpt_path, interactive=True)
-                SoVITS_dropdown = gr.Dropdown(label=i18n("SoVITS模型列表"), choices=sorted(
-                    SoVITS_names, key=custom_sort_key), value=sovits_path, interactive=True)
+                GPT_dropdown = gr.Dropdown(
+                    label=i18n("GPT模型列表"),
+                    choices=sorted(GPT_names, key=custom_sort_key),
+                    value=gpt_path,
+                    interactive=True,
+                )
+                SoVITS_dropdown = gr.Dropdown(
+                    label=i18n("SoVITS模型列表"),
+                    choices=sorted(SoVITS_names, key=custom_sort_key),
+                    value=sovits_path,
+                    interactive=True,
+                )
                 refresh_button = gr.Button(i18n("刷新模型路径"), variant="primary")
-                refresh_button.click(fn=change_choices, inputs=[], outputs=[
-                    SoVITS_dropdown, GPT_dropdown])
+                refresh_button.click(
+                    fn=change_choices,
+                    inputs=[],
+                    outputs=[SoVITS_dropdown, GPT_dropdown],
+                )
 
         with gr.Row():
             with gr.Column():
                 gr.Markdown(value=i18n("*请上传并填写参考信息"))
                 with gr.Row():
-                    inp_ref = gr.Audio("./ref.wav", type="filepath")
-                    inp_refs = gr.File(label=i18n(
-                        "辅参考音频(可选多个，或不选)"), file_count="multiple")
+                    inp_ref = gr.Audio(
+                        label="上傳 3-10 秒長嘅參考音頻", type="filepath"
+                    )
+                    inp_refs = gr.File(
+                        label=i18n("辅参考音频(可选多个，或不选)"),
+                        file_count="multiple",
+                    )
                 prompt_text = gr.Textbox(
-                    label=i18n("主参考音频的文本"), value="", lines=2)
+                    label=i18n("主参考音频的文本"), value="", lines=2
+                )
                 with gr.Row():
                     prompt_language = gr.Dropdown(
-                        label=i18n("主参考音频的语种"), choices=list(dict_language.keys()), value=i18n("中文")
+                        label=i18n("主参考音频的语种"),
+                        choices=list(dict_language.keys()),
+                        value=i18n("中文"),
                     )
                     with gr.Column():
-                        ref_text_free = gr.Checkbox(label=i18n(
-                            "开启无参考文本模式。不填参考文本亦相当于开启。"), value=False, interactive=True, show_label=True)
+                        ref_text_free = gr.Checkbox(
+                            label=i18n(
+                                "开启无参考文本模式。不填参考文本亦相当于开启。"
+                            ),
+                            value=False,
+                            interactive=True,
+                            show_label=True,
+                        )
                         gr.Markdown(
-                            i18n("使用无参考文本模式时建议使用微调的GPT，听不清参考音频说的啥(不晓得写啥)可以开，开启后无视填写的参考文本。"))
+                            i18n(
+                                "使用无参考文本模式时建议使用微调的GPT，听不清参考音频说的啥(不晓得写啥)可以开，开启后无视填写的参考文本。"
+                            )
+                        )
 
             with gr.Column():
                 gr.Markdown(value=i18n("*请填写需要合成的目标文本和语种模式"))
-                text = gr.Textbox(label=i18n("需要合成的文本"),
-                                  value="", lines=20, max_lines=20)
+                text = gr.Textbox(
+                    label=i18n("需要合成的文本"), value="", lines=20, max_lines=20
+                )
                 text_language = gr.Dropdown(
-                    label=i18n("需要合成的文本的语种"), choices=list(dict_language.keys()), value=i18n("中文")
+                    label=i18n("需要合成的文本的语种"),
+                    choices=list(dict_language.keys()),
+                    value=i18n("中文"),
                 )
 
         with gr.Group():
             gr.Markdown(value=i18n("推理设置"))
             with gr.Row():
-
                 with gr.Column():
-                    batch_size = gr.Slider(minimum=1, maximum=200, step=1, label=i18n(
-                        "batch_size"), value=20, interactive=True)
-                    fragment_interval = gr.Slider(minimum=0.01, maximum=1, step=0.01, label=i18n(
-                        "分段间隔(秒)"), value=0.3, interactive=True)
+                    batch_size = gr.Slider(
+                        minimum=1,
+                        maximum=200,
+                        step=1,
+                        label=i18n("batch_size"),
+                        value=20,
+                        interactive=True,
+                    )
+                    fragment_interval = gr.Slider(
+                        minimum=0.01,
+                        maximum=1,
+                        step=0.01,
+                        label=i18n("分段间隔(秒)"),
+                        value=0.3,
+                        interactive=True,
+                    )
                     speed_factor = gr.Slider(
-                        minimum=0.6, maximum=1.65, step=0.05, label="speed_factor", value=1.0, interactive=True)
-                    top_k = gr.Slider(minimum=1, maximum=100, step=1, label=i18n(
-                        "top_k"), value=5, interactive=True)
-                    top_p = gr.Slider(minimum=0, maximum=1, step=0.05, label=i18n(
-                        "top_p"), value=1, interactive=True)
-                    temperature = gr.Slider(minimum=0, maximum=1, step=0.05, label=i18n(
-                        "temperature"), value=1, interactive=True)
-                    repetition_penalty = gr.Slider(minimum=0, maximum=2, step=0.05, label=i18n(
-                        "重复惩罚"), value=1.35, interactive=True)
+                        minimum=0.6,
+                        maximum=1.65,
+                        step=0.05,
+                        label="speed_factor",
+                        value=1.0,
+                        interactive=True,
+                    )
+                    top_k = gr.Slider(
+                        minimum=1,
+                        maximum=100,
+                        step=1,
+                        label=i18n("top_k"),
+                        value=5,
+                        interactive=True,
+                    )
+                    top_p = gr.Slider(
+                        minimum=0,
+                        maximum=1,
+                        step=0.05,
+                        label=i18n("top_p"),
+                        value=1,
+                        interactive=True,
+                    )
+                    temperature = gr.Slider(
+                        minimum=0,
+                        maximum=1,
+                        step=0.05,
+                        label=i18n("temperature"),
+                        value=1,
+                        interactive=True,
+                    )
+                    repetition_penalty = gr.Slider(
+                        minimum=0,
+                        maximum=2,
+                        step=0.05,
+                        label=i18n("重复惩罚"),
+                        value=1.35,
+                        interactive=True,
+                    )
                 with gr.Column():
                     with gr.Row():
                         how_to_cut = gr.Dropdown(
                             label=i18n("怎么切"),
-                            choices=[i18n("不切"), i18n("凑四句一切"), i18n("凑50字一切"), i18n(
-                                "按中文句号。切"), i18n("按英文句号.切"), i18n("按标点符号切"), ],
+                            choices=[
+                                i18n("不切"),
+                                i18n("凑四句一切"),
+                                i18n("凑50字一切"),
+                                i18n("按中文句号。切"),
+                                i18n("按英文句号.���"),
+                                i18n("按标点符号切"),
+                            ],
                             value=i18n("凑四句一切"),
-                            interactive=True, scale=1
+                            interactive=True,
+                            scale=1,
+                        )
+                        parallel_infer = gr.Checkbox(
+                            label=i18n("并行推理"),
+                            value=True,
+                            interactive=True,
+                            show_label=True,
+                        )
+                        split_bucket = gr.Checkbox(
+                            label=i18n("数据分桶(并行推理时会降低一点计算量)"),
+                            value=True,
+                            interactive=True,
+                            show_label=True,
                         )
-                        parallel_infer = gr.Checkbox(label=i18n(
-                            "并行推理"), value=True, interactive=True, show_label=True)
-                        split_bucket = gr.Checkbox(label=i18n(
-                            "数据分桶(并行推理时会降低一点计算量)"), value=True, interactive=True, show_label=True)
 
                     with gr.Row():
                         seed = gr.Number(label=i18n("随机种子"), value=-1)
-                        keep_random = gr.Checkbox(label=i18n(
-                            "保持随机"), value=True, interactive=True, show_label=True)
+                        keep_random = gr.Checkbox(
+                            label=i18n("保持随机"),
+                            value=True,
+                            interactive=True,
+                            show_label=True,
+                        )
 
                     output = gr.Audio(label=i18n("输出的语音"))
                     with gr.Row():
                         inference_button = gr.Button(
-                            i18n("合成语音"), variant="primary")
+                            i18n("合成语音"), variant="primary"
+                        )
                         stop_infer = gr.Button(i18n("终止合成"), variant="primary")
 
             inference_button.click(
                 inference,
                 [
-                    text, text_language, inp_ref, inp_refs,
-                    prompt_text, prompt_language,
-                    top_k, top_p, temperature,
-                    how_to_cut, batch_size,
-                    speed_factor, ref_text_free,
-                    split_bucket, fragment_interval,
-                    seed, keep_random, parallel_infer,
-                    repetition_penalty
+                    text,
+                    text_language,
+                    inp_ref,
+                    inp_refs,
+                    prompt_text,
+                    prompt_language,
+                    top_k,
+                    top_p,
+                    temperature,
+                    how_to_cut,
+                    batch_size,
+                    speed_factor,
+                    ref_text_free,
+                    split_bucket,
+                    fragment_interval,
+                    seed,
+                    keep_random,
+                    parallel_infer,
+                    repetition_penalty,
                 ],
                 [output, seed],
             )
             stop_infer.click(tts_pipeline.stop, [], [])
-            SoVITS_dropdown.change(change_sovits_weights, [SoVITS_dropdown, prompt_language, text_language], [
-                prompt_language, text_language, prompt_text, prompt_language, text, text_language])
-            GPT_dropdown.change(
-                tts_pipeline.init_t2s_weights, [GPT_dropdown], [])
+            SoVITS_dropdown.change(
+                change_sovits_weights,
+                [SoVITS_dropdown, prompt_language, text_language],
+                [
+                    prompt_language,
+                    text_language,
+                    prompt_text,
+                    prompt_language,
+                    text,
+                    text_language,
+                ],
+            )
+            GPT_dropdown.change(tts_pipeline.init_t2s_weights, [GPT_dropdown], [])
 
         # with gr.Group():
         #     gr.Markdown(value=i18n(
@@ -348,7 +484,7 @@ async def create_app():
     return app
 
 
-if __name__ == '__main__':
+if __name__ == "__main__":
     app = asyncio.run(create_app())
     app.launch(
         # server_name="0.0.0.0",