Question

I have a thread that takes screen shots using java.awt.Robot and then encodes them into a video using Xuggler in a loop.
The loop encodes the image, then makes the thread sleep for some time depending upon the frame rate.
All good so far. The problems arise when I try to encode audio.

Specifically, maintaining the sample rate and size

I am using TargetDataLine to read data into a byte[]. This data is already BigEndian formatted.
The magic is in providing proper amount of data at proper time.
My AudioFormat looks like this:

Sample Rate: 44000Hz

Sample Size In Bits: 16

Signed: true

BigEndian: true

Assuming 10fps and 44000Hz sample rate, I will need to provide

what should be the size of the byte[]?

how much data? (measured in short because that is what Xuggler wants)

and at what time do I call the encodeAudio() method? I mean after 10 passes of the loop or 5 passes, etc.

Misc:

Community member Alex I gave me this formula:

shortArray.length == ((timeStamp - lastTimeStamp) / 1e+9) * sampleRate * channels;  

a rough calculation got me the answer of 4782 shorts for one second.
I know when you pass the audio to be encoded it must be for one full second
So I must capture 480 shorts per pass, and then encode it finally after the 10th pass.
Please tell me if this deduction is correct?

Answer 1

If you have 16 bits (two bytes) at a sample rate of 44,000 Hz that is 88,000 bytes per second. If you have stereo it is double that. In 1/10 second you need a 1/10th of that. i.e. 8,800 bytes per deci-second (1/10th of a second)

How do i retrieve short[] from ByteBuffer that has a byte[] as backing array

byte[] bytes = { };
ByteBuffer bb = ByteBuffer.wrap(bytes);

short[] shorts = new short[bb.remaining()/2];
bb.order(ByteOrder.LITTLE_ENDIAN).asShortBuffer().get(shorts);

If the order is BigEndian you don’t need to change it.