whisper.cpp python CLI

whisper.cpp-helper.py

import os
import sys
import subprocess
from pathlib import Path

LARGE_MODEL_PATH = Path("model/ggml-large.bin")
SMALL_MODEL_PATH = Path("model/ggml-small.bin")

CHECK_DIRS = ["result", "sample"]

HELP_MESSAGE = """
参数列表:
-lm --large_model: 是否使用大模型
-sm --small_model: 是否使用小模型
(默认使用小模型)
-f --ffmpeg: 使用ffmpeg进行音频转换
-nf --no-ffmpeg: 不使用ffmpeg进行音频转换
(默认使用ffmpeg进行音频转换)
-i --input <file> : 输入文件路径
-l --lang <lang> : 语言 (zh, en, ja, ko, auto)
(默认使用auto)
"""

FFMPEG_CMD_TEMPLATE = ["bin/ffmpeg.exe", "-i", "{src}", "-ar", "16000", "-ac", "1", "-c:a", "pcm_s16le", "{dst}"]
WHISPER_CMD_TEMPLATE = [
    "bin/main.exe", "-m", "{model}", "-f", "{sample}", "-t", "12", "-ocsv", "-of", "{dst}", "-l", "{lang}", "-pp", "-pc"
]

if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("参数错误")
        print(HELP_MESSAGE)
        exit(1)

    model_path = SMALL_MODEL_PATH
    use_ffmpeg = True
    lang = "auto"
    for i in range(1, len(sys.argv)):
        if sys.argv[i] == "-lm" or sys.argv[i] == "--large_model":
            model_path = LARGE_MODEL_PATH
        elif sys.argv[i] == "-sm" or sys.argv[i] == "--small_model":
            model_path = SMALL_MODEL_PATH
        elif sys.argv[i] == "-i" or sys.argv[i] == "--input":
            input_file = Path(sys.argv[i + 1])
        elif sys.argv[i] == "-f" or sys.argv[i] == "--ffmpeg":
            use_ffmpeg = True
        elif sys.argv[i] == "-nf" or sys.argv[i] == "--no-ffmpeg":
            use_ffmpeg = False
        elif sys.argv[i] == "-l" or sys.argv[i] == "--lang":
            lang = sys.argv[i + 1]

    for dir in CHECK_DIRS:
        Path(dir).mkdir(exist_ok=True)

    sample_file = input_file
    result_file = f"result/{input_file.stem}-{model_path.stem}-lang-{lang}"

    if use_ffmpeg:
        sample_file = Path("sample") / (input_file.stem + ".wav")

    if use_ffmpeg:
        cmd_list = FFMPEG_CMD_TEMPLATE.copy()
        cmd_list[2] = input_file
        cmd_list[-1] = sample_file
        subprocess.call(cmd_list)

    print("===" * 20)
    print(f"输入文件路径: {input_file}")
    print(f"采样文件路径: {sample_file}")
    print(f"结果保存路径: {result_file}")
    print(f"模型路径: {model_path}")
    print(f"是否使用 ffmpeg 进行音频转换: {use_ffmpeg}")
    print(f"语言: {lang}")
    print("===" * 20)

    cmd_list = WHISPER_CMD_TEMPLATE.copy()
    cmd_list[2] = model_path
    cmd_list[3] = sample_file
    cmd_list[-5] = result_file
    cmd_list[-3] = lang
    subprocess.call(cmd_list)

目前还是一个硬编码玩意，等以后再完善