Test audio speedup by using soundtouch

main.py

import ctypes
import math

import numpy as np
import scipy.io.wavfile as wavfile

# 加载DLL
soundtouch_dll = ctypes.CDLL("soundtouch_dll-2.3.3/SoundTouchDLL_x64.dll")

# 定义函数原型
soundtouch_dll.soundtouch_createInstance.restype = ctypes.c_void_p
soundtouch_dll.soundtouch_destroyInstance.argtypes = [ctypes.c_void_p]
soundtouch_dll.soundtouch_setRate.argtypes = [ctypes.c_void_p, ctypes.c_float]
soundtouch_dll.soundtouch_setChannels.argtypes = [ctypes.c_void_p, ctypes.c_uint]
soundtouch_dll.soundtouch_setSampleRate.argtypes = [ctypes.c_void_p, ctypes.c_uint]
soundtouch_dll.soundtouch_putSamples_i16.argtypes = [
    ctypes.c_void_p,
    ctypes.POINTER(ctypes.c_short),
    ctypes.c_uint,
]
soundtouch_dll.soundtouch_receiveSamples_i16.argtypes = [
    ctypes.c_void_p,
    ctypes.POINTER(ctypes.c_short),
    ctypes.c_uint,
]
soundtouch_dll.soundtouch_receiveSamples_i16.restype = ctypes.c_uint
soundtouch_dll.soundtouch_putSamples.argtypes = [
    ctypes.c_void_p,
    ctypes.POINTER(ctypes.c_float),
    ctypes.c_uint,
]
soundtouch_dll.soundtouch_receiveSamples.argtypes = [
    ctypes.c_void_p,
    ctypes.POINTER(ctypes.c_float),
    ctypes.c_uint,
]
soundtouch_dll.soundtouch_receiveSamples.restype = ctypes.c_uint


def process_f32(handle, data, rate):
    output_len = math.ceil(len(data) / rate)
    # output_len = math.ceil(len(data))
    soundtouch_dll.soundtouch_putSamples(handle, data, len(data))
    # 获取处理后的音频数据
    processed_samples = (ctypes.c_float * output_len)()
    received_samples = soundtouch_dll.soundtouch_receiveSamples(
        handle, processed_samples, output_len
    )
    return received_samples


def process_i16(handle, data, rate):
    output_len = math.ceil(len(data) / rate)
    # output_len = math.ceil(len(data))
    soundtouch_dll.soundtouch_putSamples_i16(handle, data, len(data))
    # 获取处理后的音频数据
    processed_samples = (ctypes.c_int16 * output_len)()
    received_samples = soundtouch_dll.soundtouch_receiveSamples_i16(
        handle, processed_samples, output_len
    )
    return received_samples


def transform(data):
    # return np.clip(data, -32768, 32767).astype(np.int16)
    return data.astype(np.float32)


def main():
    # 假设你已经有了WAV文件的音频数据samples
    sample_rate, data = wavfile.read("origin.wav")

    # 确保数据是双声道
    if len(data.shape) != 2 or data.shape[1] != 2:
        raise ValueError("Input WAV file must be stereo (2 channels).")

    # 分离双声道数据
    left_channel = data[:, 0]
    right_channel = data[:, 1]

    # 转换为 f64 并归一化
    left_channel_transform = transform(left_channel)
    right_channel_transform = transform(right_channel)

    # 创建SoundTouch实例
    handle = soundtouch_dll.soundtouch_createInstance()

    # 设置参数
    channels = 2  # 立体声
    rate = 2.0  # 增加100%的播放速度

    soundtouch_dll.soundtouch_setRate(handle, rate)
    soundtouch_dll.soundtouch_setChannels(handle, channels)
    soundtouch_dll.soundtouch_setSampleRate(handle, sample_rate)

    left_out = process_f32(handle, left_channel_transform, rate)
    right_out = process_f32(handle, right_channel_transform, rate)

    # 合并双声道数据
    output = np.column_stack((left_out, right_out))
    # 写入WAV文件
    wavfile.write("output.wav", sample_rate, output.astype(np.int16))

    # 销毁SoundTouch实例
    soundtouch_dll.soundtouch_destroyInstance(handle)


def test_f32():
    # 测试
    handle = soundtouch_dll.soundtouch_createInstance()
    receive = np.ctypeslib.as_ctypes(np.array(range(1, 100), dtype=np.float32))
    ret = process_f32(handle, receive, 2)
    print(ret, receive[:])
    soundtouch_dll.soundtouch_destroyInstance(handle)


def test_i16():
    # 测试
    handle = soundtouch_dll.soundtouch_createInstance()
    receive = np.ctypeslib.as_ctypes(np.array(range(1, 100), dtype=np.int16))
    ret = process_i16(handle, receive, 2)
    print(ret, receive[:])
    soundtouch_dll.soundtouch_destroyInstance(handle)


test_i16()