blackteay/GPT-SoVITS
1
0
Fork 0
mirror of https://github.com/RVC-Boss/GPT-SoVITS.git synced 2025-04-05 19:41:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

bug调整

Browse Source
This commit is contained in:
Downupanddownup 2024-04-29 10:13:22 +08:00
parent fe969ab9a2
commit 371a2d7138
7 changed files with 52 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Ref_Audio_Selector/config.ini
View File

@ -1,6 +1,8 @@
# config.ini
[Base]
# 服务端口号
server_port = 9423
# 参考音频目录
reference_audio_dir = refer_audio
# 临时文件目录

8
Ref_Audio_Selector/config_param/config_manager.py
View File

@ -48,7 +48,13 @@ class ParamReadWriteManager:
def write(self, key, content):
file_path = os.path.join(self.base_dir, key + '.txt')
clean_content = content.strip()
# 确保内容是字符串类型，如果不是，转换为字符串
if not isinstance(content, str):
clean_content = str(content).strip() # 转换为字符串并移除首尾空白
else:
clean_content = content.strip()
common.write_text_to_file(clean_content, file_path)

2
Ref_Audio_Selector/config_param/config_params.py
View File

@ -3,6 +3,8 @@ import Ref_Audio_Selector.config_param.config_manager as config_manager
config = config_manager.get_config()
# [Base]
# 服务端口号
server_port = int(config.get_base('server_port'))
# 参考音频目录
reference_audio_dir = config.get_base('reference_audio_dir')
# 临时文件目录

48
Ref_Audio_Selector/ref_audio_selector_webui.py
View File

@ -279,12 +279,15 @@ def text_similarity_analysis(text_work_space_dir, text_role, slider_text_similar
similarity_dir,
slider_text_similarity_amplification_boundary)
average_similarity_file = os.path.join(similarity_dir,
f'{params.text_emotion_average_similarity_report_filename}.txt')
text_text_similarity_analysis_info = f"耗时：{time_consuming:0.1f}秒；相似度分析成功：生成目录{similarity_dir}"
except Exception as e:
logger.error("发生异常: \n%s", traceback.format_exc())
text_text_similarity_analysis_info = f"发生异常：{e}"
return i18n(text_text_similarity_analysis_info)
return i18n(text_text_similarity_analysis_info), average_similarity_file
def open_text_similarity_analysis(asr_file_path, output_dir, similarity_enlarge_boundary=0.9):
@ -579,11 +582,12 @@ def init_ui():
with gr.Accordion(label=i18n("基本信息")):
with gr.Row():
text_work_space_dir = gr.Text(label=i18n("工作目录，后续操作所生成文件都会保存在此目录下"),
value=init.text_work_space_dir_default)
text_role = gr.Text(label=i18n("角色名称"), value=init.text_role_default)
button_switch_role_and_refresh = gr.Button(i18n("切换并刷新"), variant="primary")
value=init.text_work_space_dir_default, scale=2)
text_role = gr.Text(label=i18n("角色名称"), value=init.text_role_default, scale=2)
button_switch_role_and_refresh = gr.Button(i18n("切换并刷新"), variant="primary", scale=1)
text_work_space_dir.blur(save_work_dir, [text_work_space_dir, text_role], [text_role])
text_role.blur(lambda value: rw_param.write(rw_param.role, value), [text_role], [])
gr.Markdown(value=i18n("下方为公共参数，会随着进度自动填充，无需填写"))
with gr.Row():
text_refer_audio_file_dir = gr.Text(label=i18n("参考音频所在目录"),
value=init.text_refer_audio_file_dir_default)
@ -797,17 +801,21 @@ def init_ui():
text_emotion.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
[text_whole_url])
text_emotion.blur(save_emotion_param, [text_emotion], [])
gr.Markdown(value=i18n("2.3：配置待推理文本，一句一行，不要太多，10条即可"))
gr.Markdown(value=i18n("2.3：配置待推理文本，一句一行，尽量保证文本多样性，不同情绪、不同类型的都来一点"))
with gr.Row():
text_test_content = gr.Text(label=i18n("请输入待推理文本路径"), value=init.text_test_content_default)
button_open_test_content_file = gr.Button(i18n("打开待推理文本文件"), variant="primary")
button_open_test_content_file.click(open_file, [text_test_content], [])
text_test_content.blur(lambda value: rw_param.write(rw_param.test_content_path, value), [text_test_content], [])
text_test_content.blur(lambda value: rw_param.write(rw_param.test_content_path, value),
[text_test_content], [])
gr.Markdown(value=i18n("2.4：开始批量推理，这个过程比较耗时，可以去干点别的"))
slider_request_concurrency_num = gr.Slider(minimum=1, maximum=10, step=1, label=i18n(
"请输入请求并发数，会根据此数创建对应数量的子进程并行发起推理请求"), value=init.slider_request_concurrency_num_default,
slider_request_concurrency_num = gr.Slider(minimum=1, maximum=init.slider_request_concurrency_max_num,
step=1, label=i18n(
"请输入请求并发数，会根据此数创建对应数量的子进程并行发起推理请求"),
value=init.slider_request_concurrency_num_default,
interactive=True)
slider_request_concurrency_num.change(lambda value: rw_param.write(rw_param.request_concurrency_num, value), [slider_request_concurrency_num], [])
slider_request_concurrency_num.change(lambda value: rw_param.write(rw_param.request_concurrency_num, value),
[slider_request_concurrency_num], [])
with gr.Row():
button_model_inference = gr.Button(i18n("开启批量推理"), variant="primary")
text_model_inference_info = gr.Text(label=i18n("批量推理结果"), value="", interactive=False)
@ -848,7 +856,9 @@ def init_ui():
label=i18n("文本相似度放大边界"),
value=init.slider_text_similarity_amplification_boundary_default,
interactive=True)
slider_text_similarity_amplification_boundary.change(lambda value: rw_param.write(rw_param.text_similarity_amplification_boundary, value), [slider_text_similarity_amplification_boundary], [])
slider_text_similarity_amplification_boundary.change(
lambda value: rw_param.write(rw_param.text_similarity_amplification_boundary, value),
[slider_text_similarity_amplification_boundary], [])
button_asr.click(asr, [text_work_space_dir, text_role, text_asr_audio_dir, dropdown_asr_model,
dropdown_asr_size, dropdown_asr_lang],
[text_asr_info, text_text_similarity_analysis_path])
@ -856,10 +866,6 @@ def init_ui():
button_text_similarity_analysis = gr.Button(i18n("启动文本相似度分析"), variant="primary")
text_text_similarity_analysis_info = gr.Text(label=i18n("文本相似度分析结果"), value="",
interactive=False)
button_text_similarity_analysis.click(text_similarity_analysis, [text_work_space_dir, text_role,
slider_text_similarity_amplification_boundary,
text_text_similarity_analysis_path],
[text_text_similarity_analysis_info])
gr.Markdown(value=i18n("3.3：根据相似度分析结果，重点检查最后几条是否存在复读等问题"))
with gr.Row():
text_text_similarity_result_path = gr.Text(label=i18n("文本相似度分析结果文件所在路径"),
@ -867,10 +873,17 @@ def init_ui():
interactive=True)
button_open_text_similarity_result = gr.Button(i18n("打开文本相似度分析结果文件"), variant="primary")
button_open_inference_dir = gr.Button(i18n("打开推理音频所在目录"), variant="primary")
button_text_similarity_analysis.click(text_similarity_analysis, [text_work_space_dir, text_role,
slider_text_similarity_amplification_boundary,
text_text_similarity_analysis_path],
[text_text_similarity_analysis_info,
text_text_similarity_result_path])
button_open_text_similarity_result.click(open_file, [text_text_similarity_result_path], [])
button_open_inference_dir.click(open_file, [text_inference_audio_file_dir], [])
slider_audio_text_similarity_boundary = gr.Slider(minimum=0, maximum=1, step=0.01,
label=i18n("音频文本相似度边界值"), value=0.80,
slider_audio_text_similarity_boundary = gr.Slider(minimum=0, maximum=1, step=0.001,
label=i18n("音频文本相似度边界值"), value=0.800,
interactive=True)
with gr.Row():
button_delete_ref_audio_below_boundary = gr.Button(i18n("删除音频文本相似度边界值以下的参考音频"),
@ -922,7 +935,8 @@ def init_ui():
[text_model_inference_info, text_asr_audio_dir, text_inference_audio_file_dir])
app.launch(
server_port=9423,
server_port=params.server_port,
inbrowser=True,
quiet=True,
)

4
Ref_Audio_Selector/tool/audio_inference.py
View File

@ -114,9 +114,7 @@ def safe_encode_query_params(original_url):
return encoded_url
def generate_audio_files_parallel(url_composer, text_list, emotion_list, output_dir_path, num_processes=None):
if num_processes is None:
num_processes = multiprocessing.cpu_count()
def generate_audio_files_parallel(url_composer, text_list, emotion_list, output_dir_path, num_processes=1):
# 将emotion_list均匀分成num_processes个子集
emotion_groups = np.array_split(emotion_list, num_processes)

2
Ref_Audio_Selector/tool/text_comparison/asr_text_process.py
View File

@ -157,5 +157,5 @@ def parse_arguments():
if __name__ == '__main__':
cmd = parse_arguments()
print(cmd)
# print(cmd)
process(cmd.asr_file_path, cmd.output_dir, cmd.similarity_enlarge_boundary)

9
Ref_Audio_Selector/ui_init/init_ui_param.py
View File

@ -1,4 +1,5 @@
import os
import multiprocessing
import Ref_Audio_Selector.config_param.config_params as params
import Ref_Audio_Selector.common.common as common
@ -55,6 +56,8 @@ text_emotion_default = None
text_test_content_default = None
# 请求并发数
slider_request_concurrency_num_default = 3
# 最大并发数
slider_request_concurrency_max_num = None
# -------------------第三步------------------------------
@ -108,7 +111,7 @@ def init_first():
def init_second():
global text_api_set_model_base_url_default, text_api_gpt_param_default, text_api_sovits_param_default, text_api_v2_set_gpt_model_base_url_default, text_api_v2_gpt_model_param_default
global text_api_v2_set_sovits_model_base_url_default, text_api_v2_sovits_model_param_default, text_url_default, text_text_default, text_ref_path_default
global text_ref_text_default, text_emotion_default, text_test_content_default, slider_request_concurrency_num_default
global text_ref_text_default, text_emotion_default, text_test_content_default, slider_request_concurrency_num_default, slider_request_concurrency_max_num
text_api_set_model_base_url_default = empty_default(rw_param.read(rw_param.api_set_model_base_url),
'http://localhost:9880/set_model')
@ -133,8 +136,12 @@ def init_second():
text_test_content_default = empty_default(rw_param.read(rw_param.test_content_path), params.default_test_text_path)
slider_request_concurrency_max_num = multiprocessing.cpu_count()
slider_request_concurrency_num_default = empty_default(rw_param.read(rw_param.request_concurrency_num), 3)
slider_request_concurrency_num_default = min(int(slider_request_concurrency_num_default), slider_request_concurrency_max_num)
def init_third():
global text_asr_audio_dir_default, text_text_similarity_analysis_path_default, slider_text_similarity_amplification_boundary_default, text_text_similarity_result_path_default