diff --git a/GPT_SoVITS/inference_webui.py b/GPT_SoVITS/inference_webui.py
index 9c5197a7..20c6eb0e 100644
--- a/GPT_SoVITS/inference_webui.py
+++ b/GPT_SoVITS/inference_webui.py
@@ -365,7 +365,7 @@ def merge_short_text_in_array(texts, threshold):
result[len(result) - 1] += text
return result
-def get_tts_wav(ref_wav_path, prompt_text, prompt_language, text, text_language, how_to_cut=i18n("不切"), top_k=20, top_p=0.6, temperature=0.6):
+def get_tts_wav(ref_wav_path, prompt_text, prompt_language, text, text_language, how_to_cut=i18n("不切"), top_k=5, top_p=1, temperature=1):
t0 = ttime()
prompt_language = dict_language[prompt_language]
text_language = dict_language[text_language]
@@ -591,71 +591,76 @@ def get_weights_names():
SoVITS_names, GPT_names = get_weights_names()
-with gr.Blocks(title="GPT-SoVITS WebUI") as app:
- gr.Markdown(
- value=i18n("本软件以MIT协议开源, 作者不对软件具备任何控制力, 使用软件者、传播软件导出的声音者自负全责.
如不认可该条款, 则不能使用或引用软件包内任何代码和文件. 详见根目录LICENSE.")
- )
- 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)
- 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])
- SoVITS_dropdown.change(change_sovits_weights, [SoVITS_dropdown], [])
- GPT_dropdown.change(change_gpt_weights, [GPT_dropdown], [])
- gr.Markdown(value=i18n("*请上传并填写参考信息"))
- with gr.Row():
- inp_ref = gr.Audio(label=i18n("请上传3~10秒内参考音频，超过会报错！"), type="filepath")
- prompt_text = gr.Textbox(label=i18n("参考音频的文本"), value="")
- prompt_language = gr.Dropdown(
- label=i18n("参考音频的语种"), choices=[i18n("中文"), i18n("英文"), i18n("日文"), i18n("中英混合"), i18n("日英混合"), i18n("多语种混合")], value=i18n("中文")
- )
- gr.Markdown(value=i18n("*请填写需要合成的目标文本。中英混合选中文，日英混合选日文，中日混合暂不支持，非目标语言文本自动遗弃。"))
- with gr.Row():
- text = gr.Textbox(label=i18n("需要合成的文本"), value="")
- text_language = gr.Dropdown(
- label=i18n("需要合成的语种"), choices=[i18n("中文"), i18n("英文"), i18n("日文"), i18n("中英混合"), i18n("日英混合"), i18n("多语种混合")], value=i18n("中文")
- )
- how_to_cut = gr.Radio(
- 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"), ],
- value=i18n("凑四句一切"),
- interactive=True,
- )
- with gr.Row():
- 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)
- inference_button = gr.Button(i18n("合成语音"), variant="primary")
- output = gr.Audio(label=i18n("输出的语音"))
+def main():
- inference_button.click(
- get_tts_wav,
- [inp_ref, prompt_text, prompt_language, text, text_language, how_to_cut,top_k,top_p,temperature],
- [output],
+ with gr.Blocks(title="GPT-SoVITS WebUI") as app:
+ gr.Markdown(
+ value=i18n("本软件以MIT协议开源, 作者不对软件具备任何控制力, 使用软件者、传播软件导出的声音者自负全责.
如不认可该条款, 则不能使用或引用软件包内任何代码和文件. 详见根目录LICENSE.")
)
-
- gr.Markdown(value=i18n("文本切分工具。太长的文本合成出来效果不一定好，所以太长建议先切。合成会根据文本的换行分开合成再拼起来。"))
- with gr.Row():
- text_inp = gr.Textbox(label=i18n("需要合成的切分前文本"), value="")
- button1 = gr.Button(i18n("凑四句一切"), variant="primary")
- button2 = gr.Button(i18n("凑50字一切"), variant="primary")
- button3 = gr.Button(i18n("按中文句号。切"), variant="primary")
- button4 = gr.Button(i18n("按英文句号.切"), variant="primary")
- button5 = gr.Button(i18n("按标点符号切"), variant="primary")
- text_opt = gr.Textbox(label=i18n("切分后文本"), value="")
- button1.click(cut1, [text_inp], [text_opt])
- button2.click(cut2, [text_inp], [text_opt])
- button3.click(cut3, [text_inp], [text_opt])
- button4.click(cut4, [text_inp], [text_opt])
- button5.click(cut5, [text_inp], [text_opt])
- gr.Markdown(value=i18n("后续将支持混合语种编码文本输入。"))
-
-app.queue(concurrency_count=511, max_size=1022).launch(
- server_name="0.0.0.0",
- inbrowser=True,
- share=is_share,
- server_port=infer_ttswebui,
- quiet=True,
-)
+ 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)
+ 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])
+ SoVITS_dropdown.change(change_sovits_weights, [SoVITS_dropdown], [])
+ GPT_dropdown.change(change_gpt_weights, [GPT_dropdown], [])
+ gr.Markdown(value=i18n("*请上传并填写参考信息"))
+ with gr.Row():
+ inp_ref = gr.Audio(label=i18n("请上传3~10秒内参考音频，超过会报错！"), type="filepath")
+ prompt_text = gr.Textbox(label=i18n("参考音频的文本"), value="")
+ prompt_language = gr.Dropdown(
+ label=i18n("参考音频的语种"), choices=[i18n("中文"), i18n("英文"), i18n("日文"), i18n("中英混合"), i18n("日英混合"), i18n("多语种混合")], value=i18n("中文")
+ )
+ gr.Markdown(value=i18n("*请填写需要合成的目标文本。中英混合选中文，日英混合选日文，中日混合暂不支持，非目标语言文本自动遗弃。"))
+ with gr.Row():
+ text = gr.Textbox(label=i18n("需要合成的文本"), value="")
+ text_language = gr.Dropdown(
+ label=i18n("需要合成的语种"), choices=[i18n("中文"), i18n("英文"), i18n("日文"), i18n("中英混合"), i18n("日英混合"), i18n("多语种混合")], value=i18n("中文")
+ )
+ how_to_cut = gr.Radio(
+ 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"), ],
+ value=i18n("凑四句一切"),
+ interactive=True,
+ )
+ with gr.Row():
+ 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)
+ inference_button = gr.Button(i18n("合成语音"), variant="primary")
+ output = gr.Audio(label=i18n("输出的语音"))
+
+ inference_button.click(
+ get_tts_wav,
+ [inp_ref, prompt_text, prompt_language, text, text_language, how_to_cut,top_k,top_p,temperature],
+ [output],
+ )
+
+ gr.Markdown(value=i18n("文本切分工具。太长的文本合成出来效果不一定好，所以太长建议先切。合成会根据文本的换行分开合成再拼起来。"))
+ with gr.Row():
+ text_inp = gr.Textbox(label=i18n("需要合成的切分前文本"), value="")
+ button1 = gr.Button(i18n("凑四句一切"), variant="primary")
+ button2 = gr.Button(i18n("凑50字一切"), variant="primary")
+ button3 = gr.Button(i18n("按中文句号。切"), variant="primary")
+ button4 = gr.Button(i18n("按英文句号.切"), variant="primary")
+ button5 = gr.Button(i18n("按标点符号切"), variant="primary")
+ text_opt = gr.Textbox(label=i18n("切分后文本"), value="")
+ button1.click(cut1, [text_inp], [text_opt])
+ button2.click(cut2, [text_inp], [text_opt])
+ button3.click(cut3, [text_inp], [text_opt])
+ button4.click(cut4, [text_inp], [text_opt])
+ button5.click(cut5, [text_inp], [text_opt])
+ gr.Markdown(value=i18n("后续将支持混合语种编码文本输入。"))
+
+ app.queue(concurrency_count=511, max_size=1022).launch(
+ server_name="0.0.0.0",
+ inbrowser=True,
+ share=is_share,
+ server_port=infer_ttswebui,
+ quiet=True,
+ )
+
+if __name__ == '__main__':
+ main()
diff --git a/api.py b/api.py
index b8d584e7..720ef529 100644
--- a/api.py
+++ b/api.py
@@ -30,7 +30,7 @@ endpoint: `/`
使用执行参数指定的参考音频:
GET:
- `http://127.0.0.1:9880?text=先帝创业未半而中道崩殂，今天下三分，益州疲弊，此诚危急存亡之秋也。&text_language=zh`
+ `http://127.0.0.1:9880?text=先帝创业未半而中道崩殂，今天下三分，益州疲弊，此诚危急存亡之秋也。&text_language=中文`
POST:
```json
{
@@ -41,7 +41,7 @@ POST:
手动指定当次推理所使用的参考音频:
GET:
- `http://127.0.0.1:9880?refer_wav_path=123.wav&prompt_text=一二三。&prompt_language=zh&text=先帝创业未半而中道崩殂，今天下三分，益州疲弊，此诚危急存亡之秋也。&text_language=zh`
+ `http://127.0.0.1:9880?refer_wav_path=123.wav&prompt_text=一二三。&prompt_language=中文&text=先帝创业未半而中道崩殂，今天下三分，益州疲弊，此诚危急存亡之秋也。&text_language=中文`
POST:
```json
{
@@ -129,6 +129,7 @@ from text.cleaner import clean_text
from module.mel_processing import spectrogram_torch
from my_utils import load_audio
import config as global_config
+from inference_webui import get_tts_wav
g_config = global_config.Config()
@@ -316,82 +317,6 @@ dict_language = {
}
-def get_tts_wav(ref_wav_path, prompt_text, prompt_language, text, text_language):
- t0 = ttime()
- prompt_text = prompt_text.strip("\n")
- prompt_language, text = prompt_language, text.strip("\n")
- zero_wav = np.zeros(int(hps.data.sampling_rate * 0.3), dtype=np.float16 if is_half == True else np.float32)
- with torch.no_grad():
- wav16k, sr = librosa.load(ref_wav_path, sr=16000)
- wav16k = torch.from_numpy(wav16k)
- zero_wav_torch = torch.from_numpy(zero_wav)
- if (is_half == True):
- wav16k = wav16k.half().to(device)
- zero_wav_torch = zero_wav_torch.half().to(device)
- else:
- wav16k = wav16k.to(device)
- zero_wav_torch = zero_wav_torch.to(device)
- wav16k = torch.cat([wav16k, zero_wav_torch])
- ssl_content = ssl_model.model(wav16k.unsqueeze(0))["last_hidden_state"].transpose(1, 2) # .float()
- codes = vq_model.extract_latent(ssl_content)
- prompt_semantic = codes[0, 0]
- t1 = ttime()
- prompt_language = dict_language[prompt_language]
- text_language = dict_language[text_language]
- phones1, word2ph1, norm_text1 = clean_text(prompt_text, prompt_language)
- phones1 = cleaned_text_to_sequence(phones1)
- texts = text.split("\n")
- audio_opt = []
-
- for text in texts:
- phones2, word2ph2, norm_text2 = clean_text(text, text_language)
- phones2 = cleaned_text_to_sequence(phones2)
- if (prompt_language == "zh"):
- bert1 = get_bert_feature(norm_text1, word2ph1).to(device)
- else:
- bert1 = torch.zeros((1024, len(phones1)), dtype=torch.float16 if is_half == True else torch.float32).to(
- device)
- if (text_language == "zh"):
- bert2 = get_bert_feature(norm_text2, word2ph2).to(device)
- else:
- bert2 = torch.zeros((1024, len(phones2))).to(bert1)
- bert = torch.cat([bert1, bert2], 1)
-
- all_phoneme_ids = torch.LongTensor(phones1 + phones2).to(device).unsqueeze(0)
- bert = bert.to(device).unsqueeze(0)
- all_phoneme_len = torch.tensor([all_phoneme_ids.shape[-1]]).to(device)
- prompt = prompt_semantic.unsqueeze(0).to(device)
- t2 = ttime()
- with torch.no_grad():
- # pred_semantic = t2s_model.model.infer(
- pred_semantic, idx = t2s_model.model.infer_panel(
- all_phoneme_ids,
- all_phoneme_len,
- prompt,
- bert,
- # prompt_phone_len=ph_offset,
- top_k=config['inference']['top_k'],
- early_stop_num=hz * max_sec)
- t3 = ttime()
- # print(pred_semantic.shape,idx)
- pred_semantic = pred_semantic[:, -idx:].unsqueeze(0) # .unsqueeze(0)#mq要多unsqueeze一次
- refer = get_spepc(hps, ref_wav_path) # .to(device)
- if (is_half == True):
- refer = refer.half().to(device)
- else:
- refer = refer.to(device)
- # audio = vq_model.decode(pred_semantic, all_phoneme_ids, refer).detach().cpu().numpy()[0, 0]
- audio = \
- vq_model.decode(pred_semantic, torch.LongTensor(phones2).to(device).unsqueeze(0),
- refer).detach().cpu().numpy()[
- 0, 0] ###试试重建不带上prompt部分
- audio_opt.append(audio)
- audio_opt.append(zero_wav)
- t4 = ttime()
- print("%.3f\t%.3f\t%.3f\t%.3f" % (t1 - t0, t2 - t1, t3 - t2, t4 - t3))
- yield hps.data.sampling_rate, (np.concatenate(audio_opt, 0) * 32768).astype(np.int16)
-
-
def handle_control(command):
if command == "restart":
os.execl(g_config.python_exec, g_config.python_exec, *sys.argv)