修改方案，优先考虑 python 实现

python-subprocess/main.py

@@ -0,0 +1,41 @@
+from sysaudio.win import getDefaultLoopbackDevice
+from audio2text.gummy import getGummpyTranslator
+import pyaudiowpatch as pyaudio
+import numpy as np
+
+mic = pyaudio.PyAudio()
+loopback = getDefaultLoopbackDevice(mic)
+
+SAMP_WIDTH = pyaudio.get_sample_size(pyaudio.paInt16)
+FORMAT = pyaudio.paInt16
+CHANNELS = loopback["maxInputChannels"]
+RATE = int(loopback["defaultSampleRate"])
+CHUNK = RATE // 10
+INDEX = loopback["index"]
+
+
+RECORD_SECONDS = 20 # 监听时长(s)
+
+stream = mic.open(
+    format = FORMAT,
+    channels = CHANNELS,
+    rate = RATE,
+    input = True,
+    input_device_index = INDEX
+)
+
+translator = getGummpyTranslator(rate=RATE)
+translator.start()
+
+for i in range(0, int(RATE / CHUNK * RECORD_SECONDS)):
+    data = stream.read(CHUNK)
+    data_np = np.frombuffer(data, dtype=np.int16)
+    data_np_r = data_np.reshape(-1, CHANNELS)
+    mono_data = np.mean(data_np_r.astype(np.float32), axis=1)
+    mono_data = mono_data.astype(np.int16)
+    mono_data_bytes = mono_data.tobytes()
+    translator.send_audio_frame(mono_data_bytes)
+
+translator.stop()
+stream.stop_stream()
+stream.close()