【玩转腾讯云】【腾讯云语音合成TTS】短视频批量生成器

# coding=utf8
import base64
import math
import struct

from tencentcloud.common import credential
from tencentcloud.common.exception.tencent_cloud_sdk_exception import TencentCloudSDKException
from tencentcloud.tts.v20190823 import tts_client, models

SECRET_ID = "your-secret-id"
SECRET_KEY = "your-secret-key"

def int32_to_4byte(int32):
    byte_list = [0, 0, 0, 0]
    for i in range(4):
        base = math.pow(256, (3 - i))
        byte_list[3 - i] = int(int32 / base)
        int32 = int32 % base
    return byte_list

def request_tencent_tts(text):
    cred = credential.Credential(SECRET_ID, SECRET_KEY)
    client = tts_client.TtsClient(cred, "ap-shanghai")
    request = models.TextToVoiceRequest()
    request.Text = text
    request.SessionId = "session_id_123"
    request.ModelType = 1
    request.Volume = 0
    request.Speed = 0
    request.ProjectId = 0
    request.VoiceType = 0
    request.PrimaryLanguage = 1
    request.SampleRate = 16000
    request.Codec = "wav"
    response = client.TextToVoice(request)
    audio = base64.b64decode(response.Audio)
    return audio

def tts(content_fname, wav_fname):
    tts_split_char = set(['，', '。', ';'])
    with open(content_fname, "r", encoding="utf8") as f:
        content = f.read()
        step = 110
        text = ""
        last_split = 0
        last_punctuation = 0
        all_wav_pcm = b''
        audio_list = []
        for i in range(len(content)):
            if content[i] in tts_split_char:
                last_punctuation = i
            if i - last_split >= 110 or i == len(content) - 1:
                if content[i] in tts_split_char:
                    last_punctuation = i
                text = content[last_split:last_punctuation]
                if text != "":
                    #print("text: %s" % text)
                    last_split = last_punctuation + 1
                    audio = request_tencent_tts(text)
                    audio_list.append(audio)

        pcm = []
        head = []
        for audio in audio_list:
            print("audio length: %s" % len(audio))
            pcm = pcm + list(audio[44:])
            head = list(audio[:44])
            string = ""
            for c in head:
                string = string + " " + str(c)
            print(string)

        # 拿最后一个片段的文件头修改后拼接
        file_length = len(pcm) + 44
        byte_list = int32_to_4byte(file_length - 8)
        head[4] = byte_list[0]
        head[5] = byte_list[1]
        head[6] = byte_list[2]
        head[7] = byte_list[3]

        byte_list = int32_to_4byte(len(pcm))
        head[40] = byte_list[0]
        head[41] = byte_list[1]
        head[42] = byte_list[2]
        head[43] = byte_list[3]

        wav = head + pcm
            
        with open(wav_fname, 'wb') as f:
            for byte in wav:
                f.write(struct.pack('B', byte))

if __name__ == "__main__"    
    tts("content.txt", "content.wav")

# coding=utf8
from pydub import AudioSegment

def timestamp_format(msecond):
    sec, msec = divmod(msecond, 1000)
    minute, sec = divmod(sec, 60)
    hour, minute = divmod(minute, 60)
    timestamp_hour_min_sec = "{0:02d}:{1:02d}:{2:02d},{3}".format(hour, minute, sec, msec)
    print("timestamp_hour_min_sec: %s" % timestamp_hour_min_sec)
    return timestamp_hour_min_sec

def srt(speech_duration_msecond, content_fname, srt_fname):
    with open(content_fname, 'r') as f:
        content = f.read()
    srt_split_char = set(['，', '。', '：'])
    srt_text_list = []
    srt_timestamp_list = ['00:00:00,0']
    length = len(content)
    last_split_char_index = 0
    for i in range(length):
        if content[i] in srt_split_char:
            timestamp_msecond = int(i / length * speech_duration_msecond)
            item = content[last_split_char_index : i]
            last_split_char_index = i + 1
            srt_text_list.append(item)
            timestamp_hour_min_sec = timestamp_format(timestamp_msecond)
            srt_timestamp_list.append(timestamp_hour_min_sec)

    with open(srt_fname, 'w', encoding='utf8') as f:
        for i in range(len(srt_text_list)):
            item_timestamp_begin = srt_timestamp_list[i]
            item_timestamp_end = srt_timestamp_list[i + 1]
            string = "{}\n{} --> {}\n{}\n\n".format(i + 1, item_timestamp_begin, item_timestamp_end, srt_text_list[i])
            f.write(string)

def get_wav_duration(fname):
    sound= AudioSegment.from_wav(fname)
    duration = sound.duration_seconds * 1000  # 音频时长（ms）
    return duration

if __name__ == "__main__":
    duration = get_wav_duration('content.wav')
    print("duration: %d" % duration)
    srt(duration, 'content.txt', 'content.srt')

ffmpeg -i content.srt content.ass

ffmpeg -threads 2 -y -loop 1 -i pic.jpg -r 25 -i content.wav -vf ass=content.ass output.mp4