convert some text to audio using kokoro

text2audio.py

import os.path as osp
import shutil
import tempfile
import subprocess
import shlex
import argparse

SUPPORTED_EXTS = [".wav", ".mp3", ".m4b"]

print("Checking for ffmpeg...")
try:
    subprocess.check_call(
        ["ffmpeg", "--help"], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL
    )
except FileNotFoundError:
    print("ffmpeg not available. please install. exiting...")
    exit(1)

print("  ffmpeg available!")

print("Checking for kokoro and soundfile...")
try:
    from kokoro import KPipeline
    import soundfile as sf
except ImportError:
    print("kokoro or soundfile missing. please install (python modules). exiting...")
    exit(1)

print("  kokoro and soundfile available!")


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("input", help="Path to a .txt file containing the target text.")
    parser.add_argument("output", help="Output file path (has to end in .wav)")
    parser.add_argument(
        "--voice", choices=["af_heart", "af_bella", "af_nicole"], default="af_heart"
    )
    args = parser.parse_args()

    if not args.input.lower().endswith(".txt"):
        parser.error("Input file has to end on .txt")

    _, outext = osp.splitext(args.output)
    if outext.lower() not in SUPPORTED_EXTS:
        parser.error(f"Output file path has invalid suffix. Allowed: {SUPPORTED_EXTS}")

    print("Reading input text...")
    with open(args.input) as fh:
        text = fh.read()

    print("Constructing pipeline...")
    pipeline = KPipeline(lang_code="a")
    generator = pipeline(text, voice=args.voice, speed=1.25, split_pattern=r"\n+")

    print("Generating individual sound files...")
    with tempfile.TemporaryDirectory() as tmpdir:
        paths = []
        for i, (_, _, audio) in enumerate(generator):
            print(i, end=" ", flush=True)
            path = osp.join(tmpdir, f"part_{i}.wav")
            paths.append(path)
            sf.write(path, audio, 24000)
        print()

        filelist = [f"file '{fn}'\n" for fn in paths]
        filelist_path = osp.join(tmpdir, "filelist.txt")
        with open(filelist_path, "w") as fh:
            fh.writelines(filelist)

        print("Merging soundfiles...")
        merged_output = osp.join(tmpdir, "merged.wav")
        subprocess.check_call(
            shlex.split(
                f"ffmpeg -f concat -safe 0 -i {filelist_path} -c copy {merged_output}"
            )
        )

        if not outext.lower().endswith(".wav"):
            shutil.move(merged_output, args.output)
        else:
            print(f"Converting to {outext}...")
            subprocess.check_call(
                shlex.split(f"ffmpeg -i {merged_output} -q:a 2 {args.output}")
            )

    print("Done. :)")


if __name__ == "__main__":
    main()