Audio recording from microphone

[jkolobok]

I'm trying to record a pure audio file(mp3) capturing audio from the microphone and saving it using ffmpeg recorder.
I'm getting file, I can hear sound, but there is A LOT of noise. Capturing the same way but writing it using javasound api don't produce so much sound.
This is a small test. In the future I'll be creating video+audio from mic (demo shows the same problem - correct video and noisy audio).
Do you have any samples how to write pure audio?

[jkolobok]

One of the possible issues is that audio format supported by microphone is
new AudioFormat(22050,
8,
1,
signed=true,
bigEndian=true);
And if you set signed to false then I don't get any supported lines

[saudet]

You could ask this guy to see if he came up with anything:
https://code.google.com/p/javacv/issues/detail?id=493#c4
It would be great to add a sample! Let me know if you end up making one so we can add it, thanks!

[jkolobok]

I found a solution yesterday.
There is a way to transform signed audio to unsigned using
AudioSystem.getAudioInputStream(new AudioFormat(audioFormat.getSampleRate(), audioFormat.getSampleSizeInBits(),audioFormat.getChannels(), false, true), sourceStream);
where
audioFormat - original signed audio format.
sourceAudio - audio input stream that wraps targetDataLine:
AudioInputStream sourceStream = new AudioInputStream(microphoneDataLine);

i'll try to create a sample if I have time, but this should explain what to do.

[saudet]

If you're still up to creating a sample, we're still interested! Let us know what you need to get started, thanks!

[benjdavenport]

Hey dudes and dudettes,

Just wanted to chime in on this thread and share a solution to capturing CLEAN audio from a TargetDataLine source (mic or other) and writing to an FFmpegFrameRecorder instance using recordSamples (similar results can be achieved using org.bytedeco.javacv.Frame and setting the samples property).

// 44.1 sample rate, 16 bits, stereo, signed, little endian
AudioFormat audioFormat = new AudioFormat(44100.0F, 16, 2, true, false);

Mixer.Info[] minfoSet = AudioSystem.getMixerInfo();
                Mixer mixer = AudioSystem.getMixer(minfoSet[DEVICE_INDEX]);
                DataLine.Info dataLineInfo = new DataLine.Info( TargetDataLine.class, audioFormat );

// Open and start capturing audio    
                    TargetDataLine line = (TargetDataLine) mixer.getLine(dataLineInfo);
                    line.open(audioFormat);
                    line.start();

                    int audioBufferSize = (int) audioFormat.getSampleRate() * audioFormat.getChannels();
                    byte[] audioBytes = new byte[audioBufferSize];

                    AudioInputStream isS = new AudioInputStream(line);

                    while (true)
                    {
                        try
                        {
                            // Read from the line... non-blocking   
                            int nBytesRead = isS.read(audioBytes, 0, line.available());

                            if (nBytesRead > 0)
                            {
// Since we specified 16 bits in the AudioFormat, we need to convert our read byte[] to short[] (see source from FFmpegFrameRecorder.recordSamples for AV_SAMPLE_FMT_S16)
// Let's initialize our short[] array
                                int nSamplesRead = nBytesRead / 2;
                                short[] samples = new short[nSamplesRead];

// Let's wrap our short[] into a ShortBuffer and pass it to recordSamples
                             ByteBuffer.wrap(audioBytes).order(ByteOrder.LITTLE_ENDIAN).asShortBuffer().get(samples);
                                ShortBuffer sBuff = ShortBuffer.wrap(samples, 0, nSamplesRead);

                                // recorder is instance of org.bytedeco.javacv.FFmpegFrameRecorder
                                recorder.recordSamples((int) audioFormat.getSampleRate(), audioFormat.getChannels(), sBuff);
                            }
                        } 
                        catch (IOException | org.bytedeco.javacv.FrameRecorder.Exception e)
                        {
                            e.printStackTrace();
                        }
                    }

I hope this helps anyone having issues with capture / quality :o)

[saudet]

@bennyveo Nice! Thanks a lot. Would it be possible to "bundle" that into a self-contained sample, just something like a main() method with your name and message above in comments, and send a pull request to have it added to the samples directory? It would be awesome!

[benjdavenport]

Sure brutha, sounds good I’ll do that… Currently wrestling with getting timestamps in audio/video aligned, so let me tackle that and put it all in a single app. :o)

On Apr 22, 2015, at 9:38 AM, Samuel Audet [email protected] wrote:

@bennyveo https://github.com/bennyveo Nice! Thanks a lot. Would it be possible to "bundle" that into a self-contained sample, just something like a main() method with your name and message above in comments, and send a pull request to have it added to the samples directory? It would be awesome!

—
Reply to this email directly or view it on GitHub #67 (comment).

[saudet]

New sample merged: #131 Thanks @bennyveo !