.

tools/asr/config.py

@@ -1,31 +0,0 @@
-import os
-
-def check_fw_local_models():
-    '''
-    启动时检查本地是否有 Faster Whisper 模型.
-    '''
-    model_size_list = [
-        "tiny",     "tiny.en", 
-        "base",     "base.en", 
-        "small",    "small.en", 
-        "medium",   "medium.en", 
-        "large",    "large-v1", 
-        "large-v2", "large-v3"]
-    for i, size in enumerate(model_size_list):
-        if os.path.exists(f'tools/asr/models/faster-whisper-{size}'):
-            model_size_list[i] = size + '-local'
-    return model_size_list
-
-asr_dict = {
-    "达摩 ASR (中文)": {
-        'lang': ['zh'],
-        'size': ['large'],
-        'path': 'funasr_asr.py',
-    },
-    "Faster Whisper (多语种)": {
-        'lang': ['auto', 'zh', 'en', 'ja'],
-        'size': check_fw_local_models(),
-        'path': 'fasterwhisper_asr.py'
-    }
-}
-

tools/asr/fasterwhisper_asr.py

@@ -1,114 +0,0 @@
-import argparse
-import os
-import traceback
-
-os.environ["HF_ENDPOINT"]          = "https://hf-mirror.com"
-os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
-
-import torch
-from faster_whisper import WhisperModel
-from tqdm import tqdm
-
-from tools.asr.config import check_fw_local_models
-
-language_code_list = [
-    "af", "am", "ar", "as", "az", 
-    "ba", "be", "bg", "bn", "bo", 
-    "br", "bs", "ca", "cs", "cy", 
-    "da", "de", "el", "en", "es", 
-    "et", "eu", "fa", "fi", "fo", 
-    "fr", "gl", "gu", "ha", "haw", 
-    "he", "hi", "hr", "ht", "hu", 
-    "hy", "id", "is", "it", "ja", 
-    "jw", "ka", "kk", "km", "kn", 
-    "ko", "la", "lb", "ln", "lo", 
-    "lt", "lv", "mg", "mi", "mk", 
-    "ml", "mn", "mr", "ms", "mt", 
-    "my", "ne", "nl", "nn", "no", 
-    "oc", "pa", "pl", "ps", "pt", 
-    "ro", "ru", "sa", "sd", "si", 
-    "sk", "sl", "sn", "so", "sq", 
-    "sr", "su", "sv", "sw", "ta", 
-    "te", "tg", "th", "tk", "tl", 
-    "tr", "tt", "uk", "ur", "uz", 
-    "vi", "yi", "yo", "zh", "yue",
-    "auto"]
-
-def execute_asr(input_folder, output_folder, model_size, language, precision):
-    if '-local' in model_size:
-        model_size = model_size[:-6]
-        model_path = f'tools/asr/models/faster-whisper-{model_size}'
-    else:
-        model_path = model_size
-    if language == 'auto':
-        language = None #不设置语种由模型自动输出概率最高的语种
-    print("loading faster whisper model:",model_size,model_path)
-    device = 'cuda' if torch.cuda.is_available() else 'cpu'
-    try:
-        model = WhisperModel(model_path, device=device, compute_type=precision)
-    except:
-        return print(traceback.format_exc())
-    
-    input_file_names = os.listdir(input_folder)
-    input_file_names.sort()
-
-    output = []
-    output_file_name = os.path.basename(input_folder)
-    
-    for file_name in tqdm(input_file_names):
-        try:
-            file_path = os.path.join(input_folder, file_name)
-            segments, info = model.transcribe(
-                audio          = file_path,
-                beam_size      = 5,
-                vad_filter     = True,
-                vad_parameters = dict(min_silence_duration_ms=700),
-                language       = language)
-            text = ''
-
-            if info.language == "zh":
-                print("检测为中文文本, 转 FunASR 处理")
-                if("only_asr"not in globals()):
-                    from tools.asr.funasr_asr import \
-                        only_asr  # #如果用英文就不需要导入下载模型
-                text = only_asr(file_path)
-
-            if text == '':
-                for segment in segments:
-                    text += segment.text
-            output.append(f"{file_path}|{output_file_name}|{info.language.upper()}|{text}")
-        except:
-            print(traceback.format_exc())
-    
-    output_folder = output_folder or "output/asr_opt"
-    os.makedirs(output_folder, exist_ok=True)
-    output_file_path = os.path.abspath(f'{output_folder}/{output_file_name}.list')
-
-    with open(output_file_path, "w", encoding="utf-8") as f:
-        f.write("\n".join(output))
-        print(f"ASR 任务完成->标注文件路径: {output_file_path}\n")
-    return output_file_path
-
-if __name__ == '__main__':
-    parser = argparse.ArgumentParser()
-    parser.add_argument("-i", "--input_folder", type=str, required=True,
-                        help="Path to the folder containing WAV files.")
-    parser.add_argument("-o", "--output_folder", type=str, required=True, 
-                        help="Output folder to store transcriptions.")
-    parser.add_argument("-s", "--model_size", type=str, default='large-v3', 
-                        choices=check_fw_local_models(),
-                        help="Model Size of Faster Whisper")
-    parser.add_argument("-l", "--language", type=str, default='ja',
-                        choices=language_code_list,
-                        help="Language of the audio files.")
-    parser.add_argument("-p", "--precision", type=str, default='float16', choices=['float16','float32'],
-                        help="fp16 or fp32")
-
-    cmd = parser.parse_args()
-    output_file_path = execute_asr(
-        input_folder  = cmd.input_folder,
-        output_folder = cmd.output_folder,
-        model_size    = cmd.model_size,
-        language      = cmd.language,
-        precision     = cmd.precision,
-    )

tools/asr/funasr_asr.py

@@ -1,77 +0,0 @@
-# -*- coding:utf-8 -*-
-
-import argparse
-import os
-import traceback
-from tqdm import tqdm
-
-from funasr import AutoModel
-
-path_asr  = 'tools/asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'
-path_vad  = 'tools/asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch'
-path_punc = 'tools/asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch'
-path_asr  = path_asr  if os.path.exists(path_asr)  else "iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
-path_vad  = path_vad  if os.path.exists(path_vad)  else "iic/speech_fsmn_vad_zh-cn-16k-common-pytorch"
-path_punc = path_punc if os.path.exists(path_punc) else "iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch"
-
-model = AutoModel(
-    model               = path_asr,
-    model_revision      = "v2.0.4",
-    vad_model           = path_vad,
-    vad_model_revision  = "v2.0.4",
-    punc_model          = path_punc,
-    punc_model_revision = "v2.0.4",
-)
-
-def only_asr(input_file):
-    try:
-        text = model.generate(input=input_file)[0]["text"]
-    except:
-        text = ''
-        print(traceback.format_exc())
-    return text
-
-def execute_asr(input_folder, output_folder, model_size, language):
-    input_file_names = os.listdir(input_folder)
-    input_file_names.sort()
-    
-    output = []
-    output_file_name = os.path.basename(input_folder)
-
-    for file_name in tqdm(input_file_names):
-        try:
-            file_path = os.path.join(input_folder, file_name)
-            text = model.generate(input=file_path)[0]["text"]
-            output.append(f"{file_path}|{output_file_name}|{language.upper()}|{text}")
-        except:
-            print(traceback.format_exc())
-
-    output_folder = output_folder or "output/asr_opt"
-    os.makedirs(output_folder, exist_ok=True)
-    output_file_path = os.path.abspath(f'{output_folder}/{output_file_name}.list')
-
-    with open(output_file_path, "w", encoding="utf-8") as f:
-        f.write("\n".join(output))
-        print(f"ASR 任务完成->标注文件路径: {output_file_path}\n")
-    return output_file_path
-
-if __name__ == '__main__':
-    parser = argparse.ArgumentParser()
-    parser.add_argument("-i", "--input_folder", type=str, required=True,
-                        help="Path to the folder containing WAV files.")
-    parser.add_argument("-o", "--output_folder", type=str, required=True, 
-                        help="Output folder to store transcriptions.")
-    parser.add_argument("-s", "--model_size", type=str, default='large',
-                        help="Model Size of FunASR is Large")
-    parser.add_argument("-l", "--language", type=str, default='zh', choices=['zh'],
-                        help="Language of the audio files.")
-    parser.add_argument("-p", "--precision", type=str, default='float16', choices=['float16','float32'],
-                        help="fp16 or fp32")#还没接入
-
-    cmd = parser.parse_args()
-    execute_asr(
-        input_folder  = cmd.input_folder,
-        output_folder = cmd.output_folder,
-        model_size    = cmd.model_size,
-        language      = cmd.language,
-    )

tools/asr/sensevoice.py

@@ -0,0 +1,71 @@
+import argparse
+import os
+import re
+import traceback
+import torch
+from tqdm import tqdm
+from funasr import AutoModel
+
+model_dir = "tools/asr/models/SenseVoiceSmall"
+model_dir  = model_dir if os.path.exists(model_dir)  else "iic/SenseVoiceSmall"
+
+def execute_asr(input_folder, output_folder, language, device):
+    try:
+        model = AutoModel(model=model_dir,
+                  vad_model="fsmn-vad",
+                  vad_kwargs={"max_single_segment_time": 30000},
+                )
+    except:
+        return print(traceback.format_exc())
+    
+    input_file_names = os.listdir(input_folder)
+    input_file_names.sort()
+    output = []
+    output_file_name = os.path.basename(input_folder)
+    if device is None:
+        device = 'cuda' if torch.cuda.is_available() else 'cpu'
+    for file_name in tqdm(input_file_names):
+        try:
+            file_path = os.path.join(input_folder, file_name)
+            res = model.generate(
+                input=file_path,
+                cache={},
+                language=language.lower(), # "zn", "en", "yue", "ja", "ko", "nospeech"
+                use_itn=True,
+                batch_size_s=0, 
+                device = device
+            )[0]['text']
+            text_language = re.search(r'<([^<>]+)>', res).group(1)[1:-1].upper() if language == 'auto' else language.upper()
+            text = re.sub(r'<[^<>]*>', '', res).replace(' ', '')
+            output.append(f"{file_path}|{output_file_name}|{text_language}|{text}")
+        except:
+            print(traceback.format_exc())
+    
+    output_folder = output_folder or "output/asr_opt"
+    os.makedirs(output_folder, exist_ok=True)
+    output_file_path = os.path.abspath(f'{output_folder}/{output_file_name}.list')
+
+    with open(output_file_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(output))
+        print(f"ASR 任务完成->标注文件路径: {output_file_path}\n")
+    return output_file_path
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument("-i", "--input_folder", type=str, required=True,
+                        help="Path to the folder containing WAV files.")
+    parser.add_argument("-o", "--output_folder", type=str, required=True, 
+                        help="Output folder to store transcriptions.")
+    parser.add_argument("-l", "--language", type=str, default='auto',
+                        choices=['auto','zh','en','ja'],
+                        help="Language of the audio files.")
+    parser.add_argument("-d", "--device", type=str, default=None, choices=['cpu','cuda'],
+                        help="CPU or CUDA")
+
+    cmd = parser.parse_args()
+    output_file_path = execute_asr(
+        input_folder  = cmd.input_folder,
+        output_folder = cmd.output_folder,
+        language      = cmd.language,
+        device        = cmd.device,
+    )

webui.py

@@ -194,17 +194,15 @@ def change_tts_inference(if_tts,bert_path,cnhubert_base_path,gpu_number,gpt_path
         p_tts_inference=None
         yield i18n("TTS推理进程已关闭")
 
-from tools.asr.config import asr_dict
 def open_asr(asr_inp_dir, asr_opt_dir, asr_model, asr_model_size, asr_lang):
     global p_asr
     if(p_asr==None):
         asr_inp_dir=my_utils.clean_path(asr_inp_dir)
-        cmd = f'"{python_exec}" tools/asr/{asr_dict[asr_model]["path"]}'
+        cmd = f'"{python_exec}" tools/asr/sensevoice.py'
         cmd += f' -i "{asr_inp_dir}"'
         cmd += f' -o "{asr_opt_dir}"'
-        cmd += f' -s {asr_model_size}'
         cmd += f' -l {asr_lang}'
-        cmd += " -p %s"%("float16"if is_half==True else "float32")
+  
 
         yield "ASR任务开启：%s"%cmd,{"__type__":"update","visible":False},{"__type__":"update","visible":True}
         print(cmd)
@@ -732,33 +730,25 @@ with gr.Blocks(title="GPT-SoVITS WebUI") as app:
                     with gr.Row():
                         asr_model = gr.Dropdown(
                             label       = i18n("ASR 模型"),
-                            choices     = list(asr_dict.keys()),
+                            choices     = ['SenseVoice'],
                             interactive = True,
-                            value="达摩 ASR (中文)"
+                            value="enseVoice"
                         )
                         asr_size = gr.Dropdown(
                             label       = i18n("ASR 模型尺寸"),
-                            choices     = ["large"],
+                            choices     = ["small"],
                             interactive = True,
-                            value="large"
+                            value="small"
                         )
                         asr_lang = gr.Dropdown(
                             label       = i18n("ASR 语言设置"),
-                            choices     = ["zh"],
+                            choices     = ["auto","zh","en","ja"],
                             interactive = True,
-                            value="zh"
+                            value="auto"
                         )
                     with gr.Row():
                         asr_info = gr.Textbox(label=i18n("ASR进程输出信息"))
 
-                def change_lang_choices(key): #根据选择的模型修改可选的语言
-                    # return gr.Dropdown(choices=asr_dict[key]['lang'])
-                    return {"__type__": "update", "choices": asr_dict[key]['lang'],"value":asr_dict[key]['lang'][0]}
-                def change_size_choices(key): # 根据选择的模型修改可选的模型尺寸
-                    # return gr.Dropdown(choices=asr_dict[key]['size'])
-                    return {"__type__": "update", "choices": asr_dict[key]['size']}
-                asr_model.change(change_lang_choices, [asr_model], [asr_lang])
-                asr_model.change(change_size_choices, [asr_model], [asr_size])
                 
             gr.Markdown(value=i18n("0d-语音文本校对标注工具"))
             with gr.Row():