warren-buffett-speech/speak3.py

import subprocess
import os
import sys
import re

letter_year = sys.argv[1]
wav_path = sys.argv[2]

fn = f"./{letter_year}/bh-annual-letter-{letter_year}.txt"

output_dir = f"./{letter_year}/"
subprocess.run(['mkdir', '-p', output_dir])

with open(fn) as f:
    lines = f.readlines()
    all_text = ''.join(lines)
    sentences = re.matches(r"(\w\w+\.)[\s\n]", all_text)
    sentences = re.split(r"(\w\w+\.)[\s\n]", all_text)
    count = len(sentences)

    for (i, sentence) in enumerate(sentences):
        if (sentence[0] == '('):
            sentence = sentence[1:]
        if (sentence[-1] == ')'):
            sentence = sentence[:-1]

        subprocess.run([
            "tts",
            "--text", f"{sentence}",
            "--model_name", "tts_models/multilingual/multi-dataset/your_tts",
            "--speaker_wav", wav_path,
            #"--reference_wav", ref,
            # "--use_cuda", "0",
            "--out_path", f"{output_dir}/bh-{letter_year}-{str(i).zfill(3)}.wav",
            "--language_idx", "en"
            ])