app.py

"""
Copyright 2023 Balacoon

Voice Conversion service interactive demo
"""

import glob
import logging
import os
import socket
from typing import Dict

import gradio as gr

from vc_service_request import vc_service_request

script_dir = os.path.dirname(os.path.abspath(__file__))
counter = 0
# dict to limit max numer of usages per user without API key
# clears up on demo restart, but thats ok for now
users: Dict[str, int] = {}
FREE_TIER = 3


def get_ip(request: gr.Request) -> str:
    """
    returns user IP from request
    as in https://huggingface.co/spaces/radames/gradio-request-get-client-ip/blob/main/app.py
    """
    client_ip = request.client.host
    local_ip = socket.gethostbyname(socket.gethostbyname(""))
    headers = request.kwargs['headers']
    if headers and 'x-forwarded-for' in headers:
        x_forwarded_for = headers['x-forwarded-for']
        client_ip = x_forwarded_for.split(' ')[0] if x_forwarded_for else ""
    return client_ip


def main():
    logging.basicConfig(level=logging.INFO)

    badges = """
    <div style="display: flex">
    <span style="margin-right: 5px">

    [<img src="https://play.google.com/intl/en_us/badges/static/images/badges/en_badge_web_generic.png" width="200" height="77">](https://play.google.com/store/apps/details?id=com.app.vc)

    </span>
    </div>
    """

    with gr.Blocks() as demo:
        gr.Markdown(
            """
            <h1 align="center">Balacoon🦝 Voice Conversion</h1>


            Welcome to the live demo of Balacoon's Voice Conversion service.
            Check out our [website](https://balacoon.com/demo/#voice-conversion)
            to learn more.
            Voice Conversion allows you to transform your own voice
            into the voice of another person using just a single sample.
            For optimal results, we recommend using clean audio files in English.

            Here's how it works:

            1. Begin by recording your voice.
            2. Select an audio sample that represents the target voice you want to convert to.
            3. Click the "Convert" button and listen to the result!

            If providing your own audio files, please use WAVE PCM.
            Service works with 16kHz, 16 bit, mono audio.

            If you are interested to plug in Voice Conversion
            service into your own application, don't hesitate to get in touch with us at
            [[email protected]](mailto:[email protected])
            """
        )
        gr.Markdown(badges)
        with gr.Row():
            apikey = gr.Textbox(label="API key", placeholder="Enter API key or leave empty for a free tier")
        with gr.Row():
            apisecret = gr.Textbox(label="API secret", placeholder="Enter API secret or leave empty for a free tier")
        with gr.Row():
            with gr.Column(variant="panel"):
                src_audio_mic = gr.Audio(source="microphone", label="Record your voice")
                src_audio_file = gr.Audio(
                    source="upload", label="Or upload audio to convert"
                )

            with gr.Column(variant="panel"):
                tgt_audio_file = gr.Audio(
                    source="upload", label="Select audio with target voice"
                )
                tgt_examples_paths = glob.glob(
                    os.path.join(script_dir, "references", "*.wav")
                )
                gr.Examples(
                    tgt_examples_paths,
                    inputs=[tgt_audio_file],
                )

        with gr.Row():
            convert_btn = gr.Button("Convert")
        with gr.Row():
            result_audio = gr.Audio()

        def voice_conversion(src_from_mic_, src_from_file_, tgt_from_file_, api_key_, api_secret_, request_: gr.Request):
            """
            helper function which checks where source come from
            """
            src_ = None
            if src_from_mic_:
                src_ = src_from_mic_
            elif src_from_file_:
                src_ = src_from_file_
            tgt_ = tgt_from_file_
            if not src_ or not tgt_:
                logging.warning("source or target are not provided")
                gr.Warning("Source or target audio are not provided")
                return
            global counter, users
            ip_ = get_ip(request_)
            past_requests_ = users.get(ip_, 0)
            free_tier_ = False
            if not api_key_ or not api_secret_:
                free_tier = True
                api_key_ = os.environ["api_key"]
                api_secret_ = os.environ["api_secret"]
            if free_tier and past_requests_ >= FREE_TIER:
                gr.Warning("Free tier exceeded, please reach out to get your API key")
                logging.info(f"Free tier exceeded for {ip_}")
                return None
            logging.info(f"Total number of requests: {counter}. {past_requests_} for {ip_}")
            counter += 1
            users[ip_] = past_requests_ + 1
            return vc_service_request(src_, tgt_, api_key_, api_secret_)

        convert_btn.click(
            voice_conversion,
            inputs=[src_audio_mic, src_audio_file, tgt_audio_file, apikey, apisecret],
            outputs=result_audio,
        )

    demo.queue(concurrency_count=1).launch()


if __name__ == "__main__":
    main()