Игорь

Нашли с одним товарищем решение, возможно, кому-то будет полезно.
Правильный скрипт:

#!/usr/bin/env python3

import sys
import time
sys.path.append("..")
from flask import Flask, render_template, request as flask_request, send_from_directory
import ssl
from tinkoff.cloud.tts.v1 import tts_pb2_grpc, tts_pb2
from auth import authorization_metadata
import grpc
import os
import wave

endpoint = os.environ.get("VOICEKIT_ENDPOINT") or "api.tinkoff.ai:443"
api_key = os.environ.get("VOICEKIT_API_KEY")
secret_key = os.environ.get("VOICEKIT_SECRET_KEY")



# txt = 'текст, который мы передаем в переменную для синтеза речи.'
# spk = 'alyona'
# txt = request.form['data-text']
sample_rate = 48000
ssl.create_default_https_context = ssl._create_unverified_context
app = Flask(__name__, static_folder='/')


storage = os.path.join(app.root_path, "storage")
if not os.path.exists(os.path.join(app.root_path, storage)):
    os.makedirs(storage)


@app.route('/')
def index():
    return render_template('index.html')


@app.route('/build_request', methods=['POST'])
def route_build_request_post():
    data_text: str = flask_request.form.get('data_text')
    speaker_name = flask_request.form.get('speaker_name')

    filename = f'{time.strftime("%H-%M-%S", time.localtime())}.wav'

    # with open(os.path.join(storage, filename), "wb") as f:f.write(data_text.encode())
    with wave.open(filename, "wb") as f:
        f.setframerate(sample_rate)
        f.setnchannels(1)
        f.setsampwidth(2)

        stub = tts_pb2_grpc.TextToSpeechStub(
            grpc.secure_channel(endpoint, grpc.ssl_channel_credentials()))
        request = build_request(data_text, speaker_name)
        metadata = authorization_metadata(api_key, secret_key, "tinkoff.cloud.tts")
        responses = stub.StreamingSynthesize(request, metadata=metadata)
        for key, value in responses.initial_metadata():
            if key == "x-audio-duration-seconds":
                print("Estimated audio duration is {:.2f} seconds".format(float(value)))
                break
        for stream_response in responses:
            f.writeframes(stream_response.audio_chunk)

    return send_from_directory(storage, filename, as_attachment=True)


def build_request(text, name):
    return tts_pb2.SynthesizeSpeechRequest(
        input=tts_pb2.SynthesisInput(
            text=text
        ),
        audio_config=tts_pb2.AudioConfig(
            audio_encoding=tts_pb2.LINEAR16,
            sample_rate_hertz=sample_rate,
        ),
        voice=tts_pb2.VoiceSelectionParams(
            name=name,
        ),
    )


if __name__ == '__main__':
    app.run(host='0.0.0.0', port=3000, debug=True)