human/tts/tts_edge_http.py

#encoding = utf8
import logging
from io import BytesIO

import aiohttp
import numpy as np
import soundfile as sf
import edge_tts
import resampy

from .tts_base import TTSBase

logger = logging.getLogger(__name__)


class TTSEdgeHttp(TTSBase):
    def __init__(self, handle, voice='zh-CN-XiaoyiNeural'):
        super().__init__(handle)
        self._voice = voice
        # self._url = 'http://localhost:8082/v1/audio/speech'
        self._url = 'https://tts.mzzsfy.eu.org/v1/audio/speech'
        logger.info(f"TTSEdge init, {voice}")

    async def _on_request(self, txt: str):
        logger.info(f'TTSEdgeHttp, _on_request, txt:{txt}')
        data = {
            "model": "tts-1",
            "input": txt,
            "voice": "alloy",
            "speed": 1.0,
            "thread": 10
        }
        async with aiohttp.ClientSession() as session:
            async with session.post(self._url, json=data) as response:
                print('TTSEdgeHttp, _on_request, response:', response)
                if response.status == 200:
                    stream = BytesIO(await response.read())
                    return stream
                else:
                    byte_stream = None
                    return byte_stream, None

    async def _on_handle(self, stream, index):
        print('-------tts _on_handle')
        try:
            stream.seek(0)
            byte_stream = self.__create_bytes_stream(stream)
            print('-------tts start push chunk', index)
            self._handle.on_handle(byte_stream, index)
            stream.seek(0)
            stream.truncate()
            print('-------tts finish push chunk')

        except Exception as e:
            self._handle.on_handle(None, index)
            stream.seek(0)
            stream.truncate()
            print('-------tts finish error:', e)
        stream.close()

    def __create_bytes_stream(self, byte_stream):
        stream, sample_rate = sf.read(byte_stream)  # [T*sample_rate,] float64
        print(f'[INFO]tts audio stream {sample_rate}: {stream.shape}')
        stream = stream.astype(np.float32)

        if stream.ndim > 1:
            print(f'[WARN] audio has {stream.shape[1]} channels, only use the first.')
            stream = stream[:, 0]

        if sample_rate != self._handle.sample_rate and stream.shape[0] > 0:
            print(f'[WARN] audio sample rate is {sample_rate}, resampling into {self._handle.sample_rate}.')
            stream = resampy.resample(x=stream, sr_orig=sample_rate, sr_new=self._handle.sample_rate)

        return stream

    async def _on_close(self):
        print('TTSEdge close')
        # if self._byte_stream is not None and not self._byte_stream.closed:
        #     self._byte_stream.close()