What is the correct way of using SpeechRecognition's AudioData function?

I am trying to use sceech_recognition to translate my recorded audio data to text and so far I am able to successfully translate the recorded audio using a .wav . I am using pyaudio to record audio to a .wav file. Since pyaudio creates an audio stream, instead of writing it to a file and then using the file in speech_recognition, I want to translate the audio stream directly. Below is the implementation which should take the audio stream and then use sr.AudioData function and translate the audio.

`
FORMAT = pyaudio.paInt16
CHANNELS = 2
RATE = 44100
RECORD_SECONDS = 5
WAVE_OUTPUT_FILENAME = get_nonexistant_path("voice.wav")

p = pyaudio.PyAudio()

stream = p.open(format=FORMAT,
channels=CHANNELS,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)

print("* recording")

frames = []

for i in range(0, int(RATE / CHUNK * RECORD_SECONDS)):
data = stream.read(CHUNK)
frames.append(data) #appending all the intermediate streams to a list

print("* done recording")

r = sr.Recognizer()

audio_bytes = b''.join(frames) #converting list to a byte object
audio_source = sr.AudioData(audio_bytes, RATE, CHANNELS)

try:
print('Trying to convert')
text = r.recognize_google(audio_data=audio_source, language='en-US', show_all=True)
print(text)

except sr.UnknownValueError:
print("Could not understand")

stream.stop_stream()
stream.close()
p.terminate()

wf = wave.open(WAVE_OUTPUT_FILENAME, 'wb')
wf.setnchannels(CHANNELS)
wf.setsampwidth(p.get_sample_size(FORMAT))
wf.setframerate(RATE)
wf.writeframes(b''.join(frames))
wf.close()
`

Now, every time I run this code, I get an empty list as output.

`

• recording
• done recording
  Trying to convert
  [ ]
  `

Since I am also saving the audio stream to a .wav file for debugging purpose, when I use the same.wav file I get the correct translation.

You said "Google speech" {'alternative': [{'transcript': 'Google speech', 'confidence': 0.98762912}], 'final': True}
Can someone please tell me what is going wrong here.

Start by tracing the AudioData entry point and the recognize_google call, then compare the byte stream passed directly with the saved .wav file that succeeds. Confirm what input format AudioData expects and determine why the direct stream returns an empty result; done means the recorded stream produces the same transcription as the .wav path.