app.py

import os
import numpy as np
import gradio as gr
import assemblyai as aai
from translate import Translator
import uuid
from elevenlabs import VoiceSettings
from elevenlabs.client import ElevenLabs
from pathlib import Path
from scipy.io.wavfile import write, read
import yt_dlp

ELEVENLABS_API = os.environ.get("ELEVENLABS_API")
ASSEMBLYAI_API = os.environ.get("ASSEMBLYAI_API")

def voice_to_voice(audio_file):
    transcript = transcribe_audio(audio_file)
    if transcript.status == 'error':
        raise gr.Error(transcript.error)
    else:
        transcript = transcript.text

    list_translations = translate_text(transcript)
    generated_audio_paths = []

    for translation in list_translations:
        translated_audio_file_name = text_to_speech(translation)
        path = Path(translated_audio_file_name)
        generated_audio_paths.append(path)

    return tuple(generated_audio_paths + list_translations)

def transcribe_audio(audio_file):
    aai.settings.api_key = ASSEMBLYAI_API
    transcriber = aai.Transcriber()
    transcript = transcriber.transcribe(audio_file)
    return transcript

def translate_text(text):
    languages = ["ru", "tr", "sv", "de", "es", "ja", "id"]
    list_translations = []

    for lan in languages:
        translator = Translator(from_lang="en", to_lang=lan)
        translation = translator.translate(text)
        list_translations.append(translation)

    return list_translations

def text_to_speech(text):
    client = ElevenLabs(api_key=ELEVENLABS_API)
    response = client.text_to_speech.convert(
        voice_id="<your-voice-id>",
        optimize_streaming_latency="0",
        output_format="mp3_22050_32",
        text=text,
        model_id="eleven_multilingual_v2",
        voice_settings=VoiceSettings(
            stability=0.5,
            similarity_boost=0.8,
            style=0.5,
            use_speaker_boost=True,
        ),
    )

    save_file_path = f"{uuid.uuid4()}.mp3"
    with open(save_file_path, "wb") as f:
        for chunk in response:
            if chunk:
                f.write(chunk)

    return save_file_path

def download_audio(url):
    ydl_opts = {
        'format': 'bestaudio/best',
        'outtmpl': 'ytdl/%(title)s.%(ext)s',
        'postprocessors': [{
            'key': 'FFmpegExtractAudio',
            'preferredcodec': 'wav',
            'preferredquality': '192',
        }],
    }

    with yt_dlp.YoutubeDL(ydl_opts) as ydl:
        info_dict = ydl.extract_info(url, download=True)
        file_path = ydl.prepare_filename(info_dict).rsplit('.', 1)[0] + '.wav'
        sample_rate, audio_data = read(file_path)
        audio_array = np.asarray(audio_data, dtype=np.int16)

        return file_path

with gr.Blocks() as demo:
    gr.Markdown("## Audio Translator")
    gr.Markdown(
        """
        The API Key you need:
        [AssemblyAI API key](https://www.assemblyai.com/?utm_source=youtube&utm_medium=referral&utm_campaign=yt_mis_66)<br>
        [Elevenlabs API key](https://elevenlabs.io/)<br>
        Note: you need at least 30 minutes of a voice recording of yourself for the *Professional voice cloning. But there is also a simpler voice cloning option that only requires 30 seconds of voice recording. *Professional voice cloning is a paid feature.
        """
    )
    audio_input = gr.Audio(type="filepath", show_download_button=True)
    with gr.Accordion("Inputs by Link", open=False):
        with gr.Row():
            link = gr.Textbox(
                label="Link",
                placeholder="Paste the link here",
                interactive=True
            )
            download_button = gr.Button(
                "Download!",
                variant="primary"
            )
            download_button.click(download_audio, [link], [audio_input])
    submit = gr.Button("Submit", variant="primary")
    clear_button = gr.ClearButton(audio_input, "Clear")

    output_components = []
    languages = ["Turkish", "Swedish", "Russian", "German", "Spanish", "Japanese", "Indonesian"]
    
    for lang in languages:
        with gr.Group():
            output_components.append(gr.Audio(label=lang, interactive=False))
            output_components.append(gr.Markdown())

    submit.click(fn=voice_to_voice, inputs=audio_input, outputs=output_components, show_progress=True)

# Use a random port if the default one is unavailable
demo.launch(server_port=7860)