how to pass variable vaw file to whispercpp

[trappedinspacetime]

I'm trying to code a voice assistant. After hotword detected I record 3 seconds of a 16000khz mono wav

                  sample_rate = 16000
                  duration = 5  # Adjust the duration as needed
      
                  # Record audio
                  audio_data = sd.rec(int(sample_rate * duration), samplerate=sample_rate, channels=1, dtype=np.int16)
                  sd.wait()
      
                  # Convert audio data to a BytesIO object
                  audio_bytesio = io.BytesIO()
                  audio_bytesio.write(audio_data.tobytes())
                  audio_bytesio.seek(0)
      
                  # Define the whispercpp command
                  whispercpp_command = [
                      "whispercpp",
                      "-m", "/home/kenn/Desktop/2022-10/whisper.cpp/models/ggml-base.bin",
                      "-nt",
                      "-l", "tr",
                      "-f", "-",
                      "-otxt"
                  ]

I get error: failed to open WAV file from stdin
I don't know how to pass variable vaw file to `whispercpp

[trappedinspacetime]

If I save "in-memory file-like object wav" to "/dev/shm/temp.wav" and pass its path to whispercpp it works, why not in "in-memory file-like object wav"?

                sample_rate = 16000
                duration = 3  # Adjust the duration as needed
                #wav_file_path = "/dev/shm/temp_audio.wav"
    
                audio_data = sd.rec(int(sample_rate * duration), samplerate=sample_rate, channels=1, dtype=np.int16)
                sd.wait()
    
                # Save audio data to the temporary WAV file
                wav_io = io.BytesIO()
                # Save as WAV file
                sf.write('/dev/shm/temp.wav', audio_data, sample_rate, 'PCM_16')
            
                # Pass the WAV data to whispercpp
                whispercpp_command = [
                    "whispercpp",
                    "-m", "/home/kenn/Desktop/2022-10/whisper.cpp/models/ggml-base.bin",
                    "-nt",
                    "-l", "tr",
                    "-f", "/dev/shm/temp.wav",  # Whisper reads from standard input in this case
                    "-otxt"
                ]
    
                try:
                    process = subprocess.Popen(whispercpp_command, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
                    output, error = process.communicate(input=wav_io.getvalue())
    
                    # Handle the output as needed
                    if process.returncode != 0:
                        print(f"Subprocess ended with return code {process.returncode}")
                        print(f"stderr: {error.decode('utf-8')}")
                    else:
                        print(f"stdout: {output.decode('utf-8')}")
                except Exception as e:
                    print(f"Error occurred: {e}")
                             
                    if process.returncode != 0:
                        print(f"Subprocess ended with return code {process.returncode}")
                        print(f"stderr: {error.decode('utf-8')}")
                    else:
                        print(f"stdout: {output.decode('utf-8')}")
                except Exception as e:
                    print(f"Error occurred: {e}")

[trappedinspacetime]

ChatGPT saved my day:

In the first code snippet, when using subprocess.Popen with stdin=subprocess.PIPE, you are trying to pass the WAV data directly to the subprocess through its standard input. However, the whispercpp command seems to expect an actual file path with the -f option, not the audio data through standard input.

In the second code snippet, you save the audio data to a temporary WAV file (/dev/shm/temp.wav) and then pass the file path to the whispercpp command, which works.

If you want to use an in-memory file-like object, you can use the tempfile module to create a temporary file in memory. Here's an example:

    import tempfile
    
    # ... your existing code ...
    
    # Save audio data to a temporary in-memory WAV file
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_wav:
        sf.write(temp_wav.name, audio_data, sample_rate, 'PCM_16')
        temp_wav_path = temp_wav.name
    
    # Pass the WAV data to whispercpp
    whispercpp_command = [
        "whispercpp",
        "-m", "/home/kenn/Desktop/2022-10/whisper.cpp/models/ggml-base.bin",
        "-nt",
        "-l", "tr",
        "-f", temp_wav_path,  # Use the temporary file path
        "-otxt"
    ]
    
    # ... rest of your code ...

This way, you're creating a temporary in-memory WAV file using tempfile.NamedTemporaryFile and passing its path to the whispercpp command. This should achieve the desired behavior without the need to save the file to a location on disk explicitly.