Speech2Animation API

"""Moore_Speech2Animation gRPC Client"""
import argparse
import grpc
from protos.cv.digitalhuman import MotionTrigger_pb2, MotionTrigger_pb2_grpc
from utils.tts_grpc_client import simple_streaming_synthesize, dump_wav
import json
import numpy as np

def run_normal_client(host: str, tts_address: str):
    channel = grpc.insecure_channel(host)
    stub = MotionTrigger_pb2_grpc.MotionTriggerStub(channel)
    metadata=(("authorization", "<TOKEN>"),) #将<TOKEN>改为您获取的API密钥
    # text = "{开心:3}大[单手挥手 /]家好,{开心:2}我[指向自己 /]是穆莎。" #情绪标签为{情绪类型:程度}，程度为1，2，3，数值越大，情感强度越大。动作标签为[动作名称 /]，目前支持15个动作。
    text = "大家好,我是穆莎。" #输入文本
    animation_seq = []
    audio_chunks = []

    for meta, audio_chunk in simple_streaming_synthesize(address=tts_address, text=text, voice="SSB3002", metadata=metadata): #SSB3002为TTS音色
        if not meta is None:
            meta.text = text
            request = MotionTrigger_pb2.Meta2TriggerRequest(meta=meta, emotion="")
            response = stub.Meta2Trigger(request, metadata=metadata).viseme_seq
            result = list(map(lambda x:dict(x.viseme), response))
            animation_seq.extend(result)
        elif audio_chunk is not None:
            audio_chunks.append(audio_chunk)
    audio_data = np.concatenate(audio_chunks)
    dump_wav(audio_data, 22050, 'tmp', path='.') #输出音频
    with open('tmp.json', 'w') as fp:  #输出驱动信号，保存在json文件中
        data_ue = {
                    'face_data': animation_seq,
                    'fps': 30,
                    }
        json.dump(data_ue, fp)

if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("--host", type=str, default="")
    parser.add_argument("--tts_host", type=str, default="")
    args = parser.parse_args()

    run_normal_client(host=args.host, tts_address=args.tts_host)