time_stamp.py

from util import  FunASRModel,generate_results
from sentences_method import generate_new_sentences
from automodel_rec_to_sentences import convert_format,remove_chinese_punctuation #把automodel 返回的rec转换成以前Pipline的格式。
from automodel_rec_to_sentences import calculate_length,segment_text,split_into_words  #用来debug，检测两句是否一样长，是否有标点符号没去掉。



# 定义长文本写入函数
def write_long_txt_with_timestamp(wav_name, cut_line,hot_word,debug=True):
    Model = FunASRModel()
    model = Model.full_version()
    response = generate_results(model=model,wav_name=wav_name,hot_word=hot_word)
    if debug == True:
        # print(remove_chinese_punctuation(response[0]["text"]))
        sentences = convert_format(response,debug=True)
        sentences_length = 0
        for sentence in sentences:
            # 把英文单词作为一个汉字长度来计算。
            sentences_length += calculate_length(sentence)
        print(sentences,sentences_length,
              len(response[0]["timestamp"]))
        # 比对长度，如果不一样，说明有多余的未加入的符号。
        # 英文单词，不是按字母来算time_stamp的，而是按照单词来算time_stamp的，不管单词是不是有效。
        # 比如ablilly,koliyaal，这算两个词，占用两个time_stamp[start,end]x2
        # 因为有时候会识别出英文，所以需要让这个长度对齐。
        # print(response[0]["timestamp"])
    sentences = convert_format(response)


    print("=====")
    print(sentences)
    print("=====")
    # 拆分句子
    sentences = generate_new_sentences(sentences=sentences,cutline=cut_line)  ##这个会把一个句子中两句话分开，如果两句话的间隔超过cutline ,default =1000ms
    lines = []
    for i in sentences:
                skip = False
                start_time_list = []
                end_time_list = []
                ## start - end too long
                for j in i["ts_list"]:
                    ## 遍历start_end的元组
                    start_time_list.append(j[0])
                    end_time_list.append(j[1])
                # for index in range(len(start_time_list)-1): #这两个应该等长
                lines.append(str(i["ts_list"][0][0])+"|"+str(i["ts_list"][-1][-1])+"|"+i["text"])
                print(str(i["ts_list"][0][0])+"|"+str(i["ts_list"][-1][-1])+"|"+i["text"])
                # else:
                #     continue
    write_lines_to_file(f'./tmp/{wav_name}.txt', lines)
    return response


# 写入行到文件
def write_lines_to_file(file_path, lines):
    with open(file_path, 'w', encoding='utf-8') as file:
        for line in lines:
            file.write(line + '\n')


if __name__ == "__main__":
    write_long_txt_with_timestamp(wav_name="example",cut_line=1000,hot_word="",debug=True)