app.py

import gradio as gr
import os 
import shutil

#from huggingface_hub import snapshot_download
import numpy as np
from scipy.io import wavfile
"""
model_ids = [
    'suno/bark',
]

for model_id in model_ids:
    model_name = model_id.split('/')[-1]
    snapshot_download(model_id, local_dir=f'checkpoints/{model_name}')

from TTS.tts.configs.bark_config import BarkConfig
from TTS.tts.models.bark import Bark

#os.environ['CUDA_VISIBLE_DEVICES'] = '1'
config = BarkConfig()
model = Bark.init_from_config(config)
model.load_checkpoint(config, checkpoint_dir="checkpoints/bark", eval=True)
"""
from TTS.api import TTS
tts = TTS("tts_models/multilingual/multi-dataset/bark", gpu=True)

def infer(prompt, input_wav_file):

    print("SAVING THE AUDIO FILE TO WHERE IT BELONGS")

    # Path to your WAV file
    source_path = input_wav_file

    # Destination directory
    destination_directory = "bark_voices"

    # Extract the file name without the extension
    file_name = os.path.splitext(os.path.basename(source_path))[0]

    # Construct the full destination directory path
    destination_path = os.path.join(destination_directory, file_name)

    # Create the new directory
    os.makedirs(destination_path, exist_ok=True)

    # Move the WAV file to the new directory
    shutil.move(source_path, os.path.join(destination_path, f"{file_name}.wav"))

    """
    text = prompt

    print("SYNTHETIZING...")
    # with random speaker
    #output_dict = model.synthesize(text, config, speaker_id="random", voice_dirs=None)

    # cloning a speaker.
    # It assumes that you have a speaker file in `bark_voices/speaker_n/speaker.wav` or `bark_voices/speaker_n/speaker.npz`
    output_dict = model.synthesize(
        text, 
        config, 
        speaker_id=f"{file_name}", 
        voice_dirs="bark_voices/",
        gpu=True
    )
    
    print(output_dict)

    

    sample_rate = 24000  # Replace with the actual sample rate
    print("WRITING WAVE FILE")
    wavfile.write(
        'output.wav', 
        sample_rate, 
        output_dict['wav']
    )
    """
    
    tts.tts_to_file(text=prompt,
                file_path="output.wav",
                voice_dir="bark_voices/",
                speaker=f"{file_name}")

    # List all the files and subdirectories in the given directory
    contents = os.listdir(f"bark_voices/{file_name}")

    # Print the contents
    for item in contents:
        print(item)  

    tts_video = gr.make_waveform(audio="output.wav")
    
    return "output.wav", tts_video, gr.update(value=f"bark_voices/{file_name}/{contents[1]}", visible=True)


css = """
#col-container {max-width: 780px; margin-left: auto; margin-right: auto;}
img[src*='#center'] { 
    display: block;
    margin: auto;
}
.footer {
        margin-bottom: 45px;
        margin-top: 10px;
        text-align: center;
        border-bottom: 1px solid #e5e5e5;
    }
    .footer>p {
        font-size: .8rem;
        display: inline-block;
        padding: 0 10px;
        transform: translateY(10px);
        background: white;
    }
    .dark .footer {
        border-color: #303030;
    }
    .dark .footer>p {
        background: #0b0f19;
    }

.disclaimer {
    text-align: left;
}
.disclaimer > p {
    font-size: .8rem;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        
        gr.Markdown("""
        <h1 style="text-align: center;">Instant Voice Cloning</h1>
        <p style="text-align: center;">
        Clone any voice in less than 2 minutes with this <a href="https://tts.readthedocs.io/en/dev/models/bark.html" target="_blank">Coqui TSS + Bark</a> demo ! <br />
        Upload a clean 20 seconds WAV file of the voice you want to clone, <br />
        type your text-to-speech prompt and hit submit ! <br />
        </p>

        [![Duplicate this Space](https://huggingface.co/datasets/huggingface/badges/raw/main/duplicate-this-space-sm.svg#center)](https://huggingface.co/spaces/fffiloni/instant-TTS-Bark-cloning?duplicate=true)
            
        """)
        with gr.Row():
            with gr.Column():
                prompt = gr.Textbox(
                    label="Text to speech prompt"
                )
                
                audio_in = gr.Audio(
                    label="WAV voice to clone", 
                    type="filepath",
                    source="upload"
                )
                
                submit_btn = gr.Button("Submit")

            with gr.Column():
        
                cloned_out = gr.Audio(
                    label="Text to speech output"
                )
        
                video_out = gr.Video(
                    label = "Waveform video",
                    animate = True
                )
                
                npz_file = gr.File(
                    label = ".npz file",
                    visible = False
                )
    
    
    
        gr.Examples(
            examples = [
                [
                    "Once upon a time, in a cozy little shell, lived a friendly crab named Crabby. Crabby loved his cozy home, but he always felt like something was missing.",
                    "./examples/en_speaker_6.wav",
                ],
                [ 
                    "It was a typical afternoon in the bustling city, the sun shining brightly through the windows of the packed courtroom. Three people sat at the bar, their faces etched with worry and anxiety. ",
                    "./examples/en_speaker_9.wav",
                ],
            ],
            fn = infer,
            inputs = [
                prompt,
                audio_in
            ],
            outputs = [
                cloned_out, 
                video_out,
                npz_file
            ],
            cache_examples = True
        )
    
        gr.HTML("""
                <div class="footer">
                    <p>
                    Instant Voice Cloning Demo by 🤗 <a href="https://twitter.com/fffiloni" target="_blank">Sylvain Filoni</a>
                    </p>
                </div>
                <div class="disclaimer">
                    <h3> * DISCLAIMER </h3>
                    <p>
                        I hold no responsibility for the utilization or outcomes of audio content produced using the semantic constructs generated by this model. <br />
                        Please ensure that any application of this technology remains within legal and ethical boundaries. <br />
                        It is important to utilize this technology for ethical and legal purposes, upholding the standards of creativity and innovation.
                    </p>
                </div>
            """)
    
    submit_btn.click(
        fn = infer,
        inputs = [
            prompt,
            audio_in
        ],
        outputs = [
            cloned_out, 
            video_out,
            npz_file
        ]
    )

demo.queue(max_size=20).launch()