app.py

import gradio as gr
import numpy as np
import os
from scipy.io.wavfile import write
import tempfile
import zipfile
import shutil
from pydub import AudioSegment
from pydub.silence import split_on_silence
import nltk  # we'll use this to split into sentences
import subprocess

from bark import SAMPLE_RATE, generate_audio, preload_models

import numpy as np

from bark.generation import (
    generate_text_semantic,
    preload_models,
)
from bark.api import semantic_to_waveform
from bark import generate_audio, SAMPLE_RATE

# Preload models if necessary
preload_models()

def process_audio_files_with_logging(script, speaker, cloneFile):
    log_messages = "Starting audio processing...\n"
    sentences = script.split('\n')
    sentences = [item.strip() for item in sentences if item.strip()]
    GEN_TEMP = 0.4  # Example temperature, adjust as necessary

    temp_dir = tempfile.mkdtemp()

    for idx, sentence in enumerate(sentences):
        log_messages += f"Processing sentence {idx + 1}: {sentence}\n"
        semantic_tokens = generate_text_semantic(
            sentence,
            history_prompt=speaker,
            temp=GEN_TEMP,
            min_eos_p=0.05,
        )

        audio_array = semantic_to_waveform(semantic_tokens, history_prompt=speaker)
        filename = os.path.join(temp_dir, f"audio_{idx:02d}.wav")
        write(filename, SAMPLE_RATE, audio_array)
        log_messages += f"Generated audio for sentence {idx + 1}.\n"

    log_messages += "All sentences processed. Starting silence reduction...\n"

    # Process each file to remove or reduce silence
    for root, _, files in os.walk(temp_dir):
        with open("FreeVC/convert.txt", "w") as f:
            for file in files:
                file_path = os.path.join(root, file)
                audio = AudioSegment.from_file(file_path, format="wav")

                # Detect non-silent chunks and process
                processed_audio = process_audio_for_silence(audio, log_messages)

                # Overwrite the original file with processed audio
                processed_audio.export(file_path, format="wav")

                file_name_without_extension, file_extension = os.path.splitext(file)
                line = f"{file_name_without_extension}|{file_path}|{cloneFile[0]}\n"
                f.write(line)
                log_messages += line + "\n"
    #command = "python FreeVC/convert.py --hpfile FreeVC/configs/freevc.json --ptfile FreeVC/checkpoints/freevc.pth --txtpath FreeVC/convert.txt --outdir FreeVC/outputs/freevc"
    #subprocess.run(command, shell=True)
    log_messages += "Silence reduction complete. Zipping files...\n"

    # Zip the processed files
    zip_filename = zip_processed_files(temp_dir, log_messages)

    # Clean up the temporary directory
    shutil.rmtree(temp_dir)

    log_messages += "Processing complete. Files ready for download.\n"
    return zip_filename, log_messages

def process_audio_for_silence(audio, log_messages):
    # Parameters for silence detection
    silence_thresh = -32  # Silence threshold in dB
    min_silence_len = 1000  # Minimum length of silence to consider in ms
    keep_silence = 300  # Amount of silence to keep after the silence in ms

    # Detect non-silent chunks
    non_silent_chunks = split_on_silence(
        audio,
        min_silence_len=min_silence_len,
        silence_thresh=silence_thresh,
        keep_silence=keep_silence
    )

    # Combine the non-silent chunks back into a single audio segment
    processed_audio = AudioSegment.empty()
    for chunk in non_silent_chunks:
        processed_audio += chunk

    log_messages += "Audio processed for silence.\n"
    return processed_audio

def zip_processed_files(temp_dir, log_messages):
    zip_filename = os.path.join(tempfile.gettempdir(), "processed_audio_files.zip")
    with zipfile.ZipFile(zip_filename, 'w') as zipf:
        for root, _, files in os.walk(temp_dir):
            for file in files:
                zipf.write(os.path.join(root, file), file)

    log_messages += "Files zipped successfully.\n"
    return zip_filename

# Define the Gradio interface
interface = gr.Interface(
    fn=process_audio_files_with_logging,
    inputs=[gr.Textbox(label="Script", lines=10), gr.Dropdown(label="Speaker", choices=[("French","v2/fr_speaker_7"), ("English","v2/en_speaker_7"), ("Japanese","v2/ja_speaker_2"), ("German","v2/de_speaker_6"), ("Hindi","v2/hi_speaker_2"), ("Italian","v2/it_speaker_6"), ("Korean","v2/ko_speaker_0"), ("Polish","v2/pl_speaker_2"), ("Portuguese","v2/pt_speaker_5"), ("Russian","v2/ru_speaker_4"), ("Spanish","v2/es_speaker_0"), ("Turkish","v2/tr_speaker_1")]), gr.Files(label="clone voice")],
    outputs=[gr.File(label="Download Processed Files"), gr.Textbox(label="Log Messages", lines=20)],
    title="Audio Processing and Generation",
    description="Enter a script and select a speaker to generate and process audio files. Process logs will be displayed below."
)

interface.launch()