最新消息:20210816 当前crifan.com域名已被污染,为防止失联,请关注(页面右下角的)公众号

【已解决】后台用Flask封装百度的语音合成功能对外提供REST API接口

Flask crifan 132浏览 0评论
折腾:
【已解决】用和适在线的语言合成接口把文字转语音
期间,已经基本上搞清楚了思路和选择
用Flask去封装百度的语音合成的接口,
内部是用Python的SDK
是直接用pip去安装
且要合适的处理token过期的事情:
如果返回dict,且发现是err_no是502的话,则确定是token无效或过期
则使用refresh_token去重新刷新获得有效的token
重新再去尝试一次
然后正常的话,返回得到mp3的二进制数据
然后弄好了上面的部分,再去考虑:
如何把mp3二进制数据,保存到哪里,返回可访问的url给外部接口使用
后续封装出来的接口,倒是也可以考虑支持两种:
  • 直接返回mp3的url
  • 返回mp3的二进制数据文件
而返回的类型,可以通过输入参数指定
先去找找:
【已解决】Flask中如何保存临时文件且可以指定有效期
其中先要去封装百度的语音合成api
其中先要去获取token
要在Flask的服务端,用http请求:
python http
Requests: 让 HTTP 服务人类 — Requests 2.18.1 文档
安装 Requests — Requests 2.18.1 文档
所以去用requests
➜  server pipenv install requests
Installing requests…
Looking in indexes: 
https://pypi.python.org/simple
Collecting requests
  Using cached 
https://files.pythonhosted.org/packages/49/df/50aa1999ab9bde74656c2919d9c0c085fd2b3775fd3eca826012bef76d8c/requests-2.18.4-py2.py3-none-any.whl
Collecting idna<2.7,>=2.5 (from requests)
  Using cached 
https://files.pythonhosted.org/packages/27/cc/6dd9a3869f15c2edfab863b992838277279ce92663d334df9ecf5106f5c6/idna-2.6-py2.py3-none-any.whl
Collecting urllib3<1.23,>=1.21.1 (from requests)
  Using cached 
https://files.pythonhosted.org/packages/63/cb/6965947c13a94236f6d4b8223e21beb4d576dc72e8130bd7880f600839b8/urllib3-1.22-py2.py3-none-any.whl
Collecting chardet<3.1.0,>=3.0.2 (from requests)
  Using cached 
https://files.pythonhosted.org/packages/bc/a9/01ffebfb562e4274b6487b4bb1ddec7ca55ec7510b22e4c51f14098443b8/chardet-3.0.4-py2.py3-none-any.whl
Collecting certifi>=2017.4.17 (from requests)
  Using cached 
https://files.pythonhosted.org/packages/7c/e6/92ad559b7192d846975fc916b65f667c7b8c3a32bea7372340bfe9a15fa5/certifi-2018.4.16-py2.py3-none-any.whl
Installing collected packages: idna, urllib3, chardet, certifi, requests
Successfully installed certifi-2018.4.16 chardet-3.0.4 idna-2.6 requests-2.18.4 urllib3-1.22

Adding requests to Pipfile's [packages]…
Pipfile.lock (aa619c) out of date, updating to (b165df)…
Locking [dev-packages] dependencies…
Locking [packages] dependencies…
Updated Pipfile.lock (b165df)!
Installing dependencies from Pipfile.lock (b165df)…
  🐍   ▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉ 23/23 — 0
期间去搞清楚python3的url的encode:
【已解决】Python中给url中的字符串进行encode编码
【总结】
最后就基本上封装出来对应的:
获取token
刷新token
返回合成的语音数据
了,
而如何加到Flask的API中,对外提供访问接口,则是普通的做法,此处再去加上tmp的file的api。
最后代码如下:
config.py
# Audio Synthesis / TTS
BAIDU_API_KEY = "SNxxxxxxcaz"
BAIDU_SECRET_KEY = "47d7cxxxxx7ba"

AUDIO_TEMP_FOLDER = "tmp/audio"
app.py
import os
import io
import re
from urllib.parse import quote
import json
import uuid
import requests

################################################################################
# Global Definitions
################################################################################
"""
http://ai.baidu.com/docs#/TTS-API/top
    500    不支持输入
    501    输入参数不正确
    502    token验证失败
    503    合成后端错误
"""
BAIDU_ERR_NOT_SUPPORT_PARAM = 500
BAIDU_ERR_PARAM_INVALID = 501
BAIDU_ERR_TOKEN_INVALID = 502
BAIDU_ERR_BACKEND_SYNTHESIS_FAILED = 503

################################################################################
# Global Variables
################################################################################
log = None
app = None
"""
{
    "access_token": "24.569b3b5b470938a522ce60d2e2ea2506.2592000.1528015602.282335-11192483",
    "session_key": "9mzdDoR4p/oer6IHdcpJwlbK6tpH5rWqhjMi708ubA8vTgu1OToODZAXf7/963ZpEG7+yEdcdCxXq0Yp9VoSgFCFOSGEIA==",
    "scope": "public audio_voice_assistant_get audio_tts_post wise_adapt lebo_resource_base lightservice_public hetu_basic lightcms_map_poi kaidian_kaidian ApsMisTest_Test权限 vis-classify_flower lpq_开放 cop_helloScope ApsMis_fangdi_permission smartapp_snsapi_base",
    "refresh_token": "25.5acf5c4d9fdfdbe577e75f3f2fd137b8.315360000.1840783602.282335-11192483",
    "session_secret": "121fe91236ef88ab24b2ecab479427ea",
    "expires_in": 2592000
}
"""
gCurBaiduRespDict = {} # get baidu token resp dict
gTempAudioFolder = “"


################################################################################
# Global Function
################################################################################

def generateUUID(prefix = ""):
    generatedUuid4 = uuid.uuid4()
    generatedUuid4Str = str(generatedUuid4)
    newUuid = prefix + generatedUuid4Str
    return newUuid

#----------------------------------------
# Audio Synthesis / TTS
#----------------------------------------

def createAudioTempFolder():
    """create foler to save later temp audio files"""
    global log, gTempAudioFolder

    # init audio temp folder for later store temp audio file
    audioTmpFolder = app.config["AUDIO_TEMP_FOLDER"]
    log.info("audioTmpFolder=%s", audioTmpFolder)
    curFolderAbsPath = os.getcwd() #'/Users/crifan/dev/dev_root/company/xxx/projects/robotDemo/server'
    log.info("curFolderAbsPath=%s", curFolderAbsPath)
    audioTmpFolderFullPath = os.path.join(curFolderAbsPath, audioTmpFolder)
    log.info("audioTmpFolderFullPath=%s", audioTmpFolderFullPath)
    if not os.path.exists(audioTmpFolderFullPath):
        os.makedirs(audioTmpFolderFullPath)

    gTempAudioFolder = audioTmpFolderFullPath
    log.info("gTempAudioFolder=%s", gTempAudioFolder)

def initAudioSynthesis():
    """
    init audio synthesis related:
        init token
    :return:
    """

    getBaiduToken()
    createAudioTempFolder()


def getBaiduToken():
    """get baidu token"""
    global app, log, gCurBaiduRespDict

    getBaiduTokenUrlTemplate = "https://openapi.baidu.com/oauth/2.0/token?grant_type=client_credentials&client_id=%s&client_secret=%s"
    getBaiduTokenUrl = getBaiduTokenUrlTemplate % (app.config["BAIDU_API_KEY"], app.config["BAIDU_SECRET_KEY"])
    log.info("getBaiduTokenUrl=%s", getBaiduTokenUrl) #https://openapi.baidu.com/oauth/2.0/token?grant_type=client_credentials&client_id=SNjsxxxxcaz&client_secret=47d7c0xxx7ba
    resp = requests.get(getBaiduTokenUrl)
    log.info("resp=%s", resp)
    respJson = resp.json()
    log.info("respJson=%s", respJson) #{'access_token': '24.8f1f35xxxx5.2592000.1xxx0.282335-11192483', 'session_key': '9mzxxxp5ZUHafqq8m+6KwgZmw==', 'scope': 'public audio_voice_assistant_get audio_tts_post wise_adapt lebo_resource_base lightservice_public hetu_basic lightcms_map_poi kaidian_kaidian ApsMisTest_Test权限 vis-classify_flower lpq_开放 cop_helloScope ApsMis_fangdi_permission smartapp_snsapi_base', 'refresh_token': '25.eb2xxxe.315360000.1841377320.282335-11192483', 'session_secret': 'c0a83630b7b1c46039e360de417d346e', 'expires_in': 2592000}
    if resp.status_code == 200:
        gCurBaiduRespDict = respJson
        log.info("get baidu token resp: %s", gCurBaiduRespDict)
    else:
        log.error("error while get baidu token: %s", respJson)
        #{'error': 'invalid_client', 'error_description': 'Client authentication failed'}
        #{'error': 'invalid_client', 'error_description': 'unknown client id'}
        #{'error': 'unsupported_grant_type', 'error_description': 'The authorization grant type is not supported'}

def refreshBaiduToken():
    """refresh baidu token when current token invalid"""
    global app, log, gCurBaiduRespDict
    if gCurBaiduRespDict:
        refreshBaiduTokenUrlTemplate = "https://openapi.baidu.com/oauth/2.0/token?grant_type=refresh_token&refresh_token=%s&client_id=%s&client_secret=%s"
        refreshBaiduTokenUrl = refreshBaiduTokenUrlTemplate % (gCurBaiduRespDict["refresh_token"], app.config["BAIDU_API_KEY"], app.config["BAIDU_SECRET_KEY"])
        log.info("refreshBaiduTokenUrl=%s", refreshBaiduTokenUrl) #https://openapi.baidu.com/oauth/2.0/token?grant_type=refresh_token&refresh_token=25.1b7xxxxedbb99ea361.31536xx.1841379583.282335-11192483&client_id=SNjsggdYDNWtnlbKhxsPLcaz&client_secret=47dxxxxba
        resp = requests.get(refreshBaiduTokenUrl)
        log.info("resp=%s", resp)
        respJson = resp.json()
        log.info("respJson=%s", respJson)
        if resp.status_code == 200:
            gCurBaiduRespDict = respJson
            log.info("Ok to refresh baidu token response: %s", gCurBaiduRespDict)
        else:
            log.error("error while refresh baidu token: %s", respJson)
    else:
        log.error("Can't refresh baidu token for previous not get token")


def baiduText2Audio(unicodeText):
    """call baidu text2audio to generate mp3 audio from text"""
    global app, log, gCurBaiduRespDict
    log.info("baiduText2Audio: unicodeText=%s", unicodeText)

    isOk = False
    mp3BinData = None
    errNo = 0
    errMsg = "Unknown error"

    if not gCurBaiduRespDict:
        errMsg = "Need get baidu token before call text2audio"
        return isOk, mp3BinData, errNo, errMsg

    utf8Text = unicodeText.encode("utf-8")
    log.info("utf8Text=%s", utf8Text)
    encodedUtf8Text = quote(unicodeText)
    log.info("encodedUtf8Text=%s", encodedUtf8Text)

    # http://ai.baidu.com/docs#/TTS-API/top
    tex = encodedUtf8Text #合成的文本,使用UTF-8编码。小于512个中文字或者英文数字。(文本在百度服务器内转换为GBK后,长度必须小于1024字节)
    tok = gCurBaiduRespDict["access_token"] #开放平台获取到的开发者access_token(见上面的“鉴权认证机制”段落)
    cuid = app.config["FLASK_APP_NAME"] #用户唯一标识,用来区分用户,计算UV值。建议填写能区分用户的机器 MAC 地址或 IMEI 码,长度为60字符以内
    ctp = 1 #客户端类型选择,web端填写固定值1
    lan = "zh" #固定值zh。语言选择,目前只有中英文混合模式,填写固定值zh
    spd = 5 #语速,取值0-9,默认为5中语速
    pit = 5 #音调,取值0-9,默认为5中语调
    # vol = 5 #音量,取值0-9,默认为5中音量
    vol = 9
    per = 0 #发音人选择, 0为普通女声,1为普通男生,3为情感合成-度逍遥,4为情感合成-度丫丫,默认为普通女声
    getBaiduSynthesizedAudioTemplate = "http://tsn.baidu.com/text2audio?lan=%s&ctp=%s&cuid=%s&tok=%s&vol=%s&per=%s&spd=%s&pit=%s&tex=%s"
    getBaiduSynthesizedAudioUrl = getBaiduSynthesizedAudioTemplate % (lan, ctp, cuid, tok, vol, per, spd, pit, tex)
    log.info("getBaiduSynthesizedAudioUrl=%s", getBaiduSynthesizedAudioUrl) #http://tsn.baidu.com/text2audio?lan=zh&ctp=1&cuid=RobotQA&tok=24.5xxxb5.2592000.1528609737.282335-11192483&vol=5&per=0&spd=5&pit=5&tex=as%20a%20book-collector%2C%20i%20have%20the%20story%20you%20just%20want%20to%20listen%21
    resp = requests.get(getBaiduSynthesizedAudioUrl)
    log.info("resp=%s", resp)
    respContentType = resp.headers["Content-Type"]
    respContentTypeLowercase = respContentType.lower() #'audio/mp3'
    log.info("respContentTypeLowercase=%s", respContentTypeLowercase)
    if respContentTypeLowercase == "audio/mp3":
        mp3BinData = resp.content
        log.info("resp content is binary data of mp3, length=%d", len(mp3BinData))
        isOk = True
        errMsg = ""
    elif respContentTypeLowercase == "application/json":
        """                        
            {
              'err_detail': 'Invalid params per or lan!',
              'err_msg': 'parameter error.',
              'err_no': 501,
              'err_subcode': 50000,
              'tts_logid': 642798357
            }

            {
              'err_detail': 'Invalid params per&pdt!',
              'err_msg': 'parameter error.',
              'err_no': 501,
              'err_subcode': 50000,
              'tts_logid': 1675521246
            }

            {
              'err_detail': 'Access token invalid or no longer valid',
              'err_msg': 'authentication failed.',
              'err_no': 502,
              'err_subcode': 50004,
              'tts_logid': 4221215043
            }
        """
        log.info("resp content is json -> occur error")

        isOk = False
        respDict = resp.json()
        log.info("respDict=%s", respDict)
        errNo = respDict["err_no"]
        errMsg = respDict["err_msg"] + " " + respDict["err_detail"]
    else:
        isOk = False
        errMsg = "Unexpected response content-type: %s" % respContentTypeLowercase

    return isOk, mp3BinData, errNo, errMsg

def doAudioSynthesis(unicodeText):
    """
        do audio synthesis from unicode text
        if failed for token invalid/expired, will refresh token to do one more retry
    """
    global app, log, gCurBaiduRespDict
    isOk = False
    audioBinData = None
    errMsg = ""

    # # for debug
    # gCurBaiduRespDict["access_token"] = "99.5xxx06.2592000.1528015602.282335-11192483"

    log.info("doAudioSynthesis: unicodeText=%s", unicodeText)
    isOk, audioBinData, errNo, errMsg = baiduText2Audio(unicodeText)
    log.info("isOk=%s, errNo=%d, errMsg=%s", isOk, errNo, errMsg)

    if isOk:
        errMsg = ""
        log.info("got synthesized audio binary data length=%d", len(audioBinData))
    else:
        if errNo == BAIDU_ERR_TOKEN_INVALID:
            log.warning("Token invalid -> refresh token")
            refreshBaiduToken()

            isOk, audioBinData, errNo, errMsg = baiduText2Audio(unicodeText)
            log.info("after refresh token: isOk=%ss, errNo=%s, errMsg=%s", isOk, errNo, errMsg)
        else:
            log.warning("try synthesized audio occur error: errNo=%d, errMsg=%s", errNo, errMsg)
            audioBinData = None

    log.info("return isOk=%s, errMsg=%s", isOk, errMsg)
    if audioBinData:
        log.info("audio binary bytes=%d", len(audioBinData))
    return isOk, audioBinData, errMsg


def testAudioSynthesis():
    global app, log, gCurBaiduRespDict, gTempAudioFolder

    testInputUnicodeText = u"as a book-collector, i have the story you just want to listen!"
    isOk, audioBinData, errMsg = doAudioSynthesis(testInputUnicodeText)
    if isOk:
        audioBinDataLen = len(audioBinData)
        log.info("Now will save audio binary data %d bytes to file", audioBinDataLen)

        # 1. save mp3 binary data into tmp file
        newUuid = generateUUID()
        log.info("newUuid=%s", newUuid)
        tempFilename = newUuid + ".mp3"
        log.info("tempFilename=%s", tempFilename)
        if not gTempAudioFolder:
            createAudioTempFolder()
        tempAudioFullname = os.path.join(gTempAudioFolder, tempFilename) #'/Users/crifan/dev/dev_root/company/xxx/projects/robotDemo/server/tmp/audio/2aba73d1-f8d0-4302-9dd3-d1dbfad44458.mp3'
        log.info("tempAudioFullname=%s", tempAudioFullname)

        with open(tempAudioFullname, 'wb') as tmpAudioFp:
            log.info("tmpAudioFp=%s", tmpAudioFp)
            tmpAudioFp.write(audioBinData)
            tmpAudioFp.close()
            log.info("Done to write audio data into file of %d bytes", audioBinDataLen)

        # TODO:
        # 2. use celery to delay delete tmp file
    else:
        log.warning("Fail to get synthesis audio for errMsg=%s", errMsg)


#----------------------------------------
# Flask API
#----------------------------------------
def sendFile(fileBytes, contentType, outputFilename):
    """Flask API use this to send out file (to browser, browser can directly download file)"""
    return send_file(
        io.BytesIO(fileBytes),
        # io.BytesIO(fileObj.read()),
        mimetype=contentType,
        as_attachment=True,
        attachment_filename=outputFilename
    )

################################################################################
# Global Init App
################################################################################
app = Flask(__name__)
CORS(app)
app.config.from_object('config.DevelopmentConfig')
# app.config.from_object('config.ProductionConfig')

logFormatterStr = app.config["LOG_FORMAT"]
logFormatter = logging.Formatter(logFormatterStr)

fileHandler = RotatingFileHandler(
    app.config['LOG_FILE_FILENAME'],
    maxBytes=app.config["LOF_FILE_MAX_BYTES"],
    backupCount=app.config["LOF_FILE_BACKUP_COUNT"],
    encoding="UTF-8")
fileHandler.setLevel(logging.DEBUG)
fileHandler.setFormatter(logFormatter)
app.logger.addHandler(fileHandler)

app.logger.setLevel(logging.DEBUG) # set root log level

log = app.logger
log.debug("app=%s", app)
log.info("app.config=%s", app.config)

api = Api(app)
log.info("api=%s", api)

aiContext = Context()
log.info("aiContext=%s", aiContext)

initAudioSynthesis()
# testAudioSynthesis()


class RobotQaAPI(Resource):

    def processResponse(self, respDict):
        """
            process response dict before return
                generate audio for response text part
        """
        global log, gTempAudioFolder

        tmpAudioUrl = ""

        unicodeText = respDict["data"]["response"]["text"]
        log.info("unicodeText=%s")

        if not unicodeText:
            log.info("No response text to do audio synthesis")
            return jsonify(respDict)

        isOk, audioBinData, errMsg = doAudioSynthesis(unicodeText)
        if isOk:
            audioBinDataLen = len(audioBinData)
            log.info("audioBinDataLen=%s", audioBinDataLen)

            # 1. save mp3 binary data into tmp file
            newUuid = generateUUID()
            log.info("newUuid=%s", newUuid)
            tempFilename = newUuid + ".mp3"
            log.info("tempFilename=%s", tempFilename)
            if not gTempAudioFolder:
                createAudioTempFolder()
            tempAudioFullname = os.path.join(gTempAudioFolder, tempFilename)
            log.info("tempAudioFullname=%s", tempAudioFullname) # 'xxx/tmp/audio/2aba73d1-f8d0-4302-9dd3-d1dbfad44458.mp3'

            with open(tempAudioFullname, 'wb') as tmpAudioFp:
                log.info("tmpAudioFp=%s", tmpAudioFp)
                tmpAudioFp.write(audioBinData)
                tmpAudioFp.close()
                log.info("Saved %d bytes data into temp audio file %s", audioBinDataLen, tempAudioFullname)

            # TODO:
            # 2. use celery to delay delete tmp file

            # generate temp audio file url
            # /tmp/audio
            tmpAudioUrl = "http://%s:%d/tmp/audio/%s" % (
                app.config["FILE_URL_HOST"],
                app.config["FLASK_PORT"],
                tempFilename)
            log.info("tmpAudioUrl=%s", tmpAudioUrl)
            respDict["data"]["response"]["audioUrl"] = tmpAudioUrl
        else:
            log.warning("Fail to get synthesis audio for errMsg=%s", errMsg)

        log.info("respDict=%s", respDict)
        return jsonify(respDict)


def get(self):
    respDict = {
        "code": 200,
        "message": "generate response ok",
        "data": {
            "input": "",
            "response": {
                "text": "",
                "audioUrl": ""
            },
            "control": "",
            "audio": {}
        }
    }

    。。。
            respDict["data"]["audio"] = {
                "contentType": audioFileObj.contentType,
                "name": audioFileObj.filename,
                "size": audioFileObj.length,
                "url": "http://%s:%d/files/%s/%s" %
                       (app.config["FILE_URL_HOST"],
                        app.config["FLASK_PORT"],
                        audioFileObj._id,
                        encodedFilename)
            }
            log.info("respDict=%s", respDict)
            return self.processResponse(respDict)


class TmpAudioAPI(Resource):

    def get(self, filename=None):
        global gTempAudioFolder

        log.info("TmpAudioAPI: filename=%s", filename)

        tmpAudioFullPath = os.path.join(gTempAudioFolder, filename)
        log.info("tmpAudioFullPath=%s", tmpAudioFullPath)

        if not os.path.isfile(tmpAudioFullPath):
            log.warning("Not exists file %s", tmpAudioFullPath)
            respDict = {
                "code": 404,
                "message": "Can not find temp audio file %s" % filename,
                "data": {}
            }
            return jsonify(respDict)

        fileSize = os.path.getsize(tmpAudioFullPath)
        log.info("fileSize=%s", fileSize)

        with open(tmpAudioFullPath, "rb") as tmpAudioFp:
            fileBytes = tmpAudioFp.read()
            log.info("read out fileBytes length=%s", len(fileBytes))

            outputFilename = filename
            # contentType = "audio/mp3" # chrome use this
            contentType = "audio/mpeg" # most common and compatible

            return sendFile(fileBytes, contentType, outputFilename)

api.add_resource(RobotQaAPI, '/qa', endpoint='qa')
api.add_resource(TmpAudioAPI, '/tmp/audio/<filename>', endpoint='TmpAudio')
如此即可实现:
访问qa,返回内容中包含了,从百度的语音合成接口返回的二进制的mp3数据,保存到临时文件夹后,并返回生成的本地的Flask的tmp的file的url:
然后再去访问该url
http://127.0.0.1:32851/tmp/audio/edaf4826-8f8a-4c31-8de5-1ce2debff2e8.mp3
可以下载到mp3文件:
剩下的就是:
【未解决】Flask中如何保存临时文件且可以指定有效期
了。

转载请注明:在路上 » 【已解决】后台用Flask封装百度的语音合成功能对外提供REST API接口

发表我的评论
取消评论

表情

Hi,您需要填写昵称和邮箱!

  • 昵称 (必填)
  • 邮箱 (必填)
  • 网址
93 queries in 0.168 seconds, using 20.80MB memory