convert text to audio using balacoon library

text2audio.py

from balacoon_tts import TTS
import soundfile as sf
import numpy as np
import traceback
import os
import subprocess
import re
from icecream import ic
import sys
# >>> ============================================================ <<<
# >>> IMPORTANT: SET YOUR BALACOON MODEL ADDON FILE PATH HERE      <<<
BALACOON_MODEL_PATH = "/home/jack/hidden/en_us_cmuartic_jets_cpu.addon"  # EDIT THIS
# >>> ============================================================ <<<

tts_instance = None
selected_speaker = None
AUDIO_FOLDER = 'static/new_audio'

# ---------------- SAFE FILENAME FUNCTION -------------------------
def safe_filename(text, count):
    """Generates a safe filename from the first 5 words of text."""
    words = text.strip().split()
    first_words = "_".join(words[:5])
    # Remove unsafe characters
    cleaned = re.sub(r'[^\w\s-]', '', first_words)
    cleaned = cleaned.replace(" ", "_").replace("-", "_")
    cleaned = cleaned[:50]
    if not cleaned:
        cleaned = f"entry_{count}_audio"
    return cleaned

# ---------------- INITIALIZE TTS ----------------------------------
ic(f"Loading Balacoon model from: {BALACOON_MODEL_PATH}")
tts_instance = TTS(BALACOON_MODEL_PATH)
ic("Balacoon TTS instance created.")

supported_speakers = tts_instance.get_speakers()
ic(supported_speakers)
# speaker -3 male nice voice
selected_speaker = supported_speakers[-3]
ic(f"Automatically selected speaker: '{selected_speaker}'")

# ---------------- FUNCTION TO GENERATE AUDIO --------------------
def generate_audio_file(text, count, filename):
    """Generates WAV and MP3 audio files, returns MP3 path."""
    if not tts_instance or not selected_speaker:
        ic("Error: Balacoon TTS instance or speaker not available.")
        return None, "Balacoon TTS not initialized correctly."
    try:
        os.makedirs(AUDIO_FOLDER, exist_ok=True)
        base_filename = f"{filename}"
        wav_path = os.path.join(AUDIO_FOLDER, base_filename + ".wav")
        mp3_path = os.path.join(AUDIO_FOLDER, base_filename + ".mp3")
        ic(wav_path, mp3_path)
        text = text.replace("*", "").replace("…", ", ")
        ic("Synthesizing WAV...")
        audio_data = tts_instance.synthesize(text=text, speaker=selected_speaker)
        ic(audio_data.shape, audio_data.dtype)
        # Save WAV
        balacoon_samplerate = 24000
        sf.write(wav_path, audio_data, balacoon_samplerate)
        ic(f"WAV saved: {wav_path}")
        # Convert to MP3 using ffmpeg
        ic("Converting to MP3...")
        command = [
            "ffmpeg",
            "-y",
            "-i", wav_path,
            "-codec:a", "libmp3lame",
            "-qscale:a", "2",
            mp3_path
        ]
        subprocess.run(command, check=True)
        ic(f"MP3 created: {mp3_path}")
        # Remove the .wav to save space
        os.remove(wav_path)
        ic(f"Deleted temp WAV: {wav_path}")
        static_path = os.path.join("audio", filename + ".mp3").replace("\\", "/")
        return static_path, None

    except Exception as e:
        error_message = f"Error generating audio for ID {count}: {e}"
        ic(error_message)
        traceback.print_exc()
        return None, error_message

# ------------------- READ TEXT FILE ------------------------------
def read_text(text_file):
    count = 0
    with open(text_file, 'r', encoding='utf-8') as f:
        content = f.read()
    blocks = content.split("\n\n\n")
    for text in blocks:
        count += 1
        ic(count, text[:60])
        filename = safe_filename(text, count)
        ic(f"Using filename: {filename}")
        generate_audio_file(text, count, filename)

# ------------------- MAIN EXECUTION BLOCK -----------------------
if __name__ == '__main__':
    text_file = sys.argv[1]#'static/TXT/'
    read_text(text_file)