Encoding Raw Data to Flac

[mertakdut]

Hello,

Getting an error while trying to encode raw data which I decoded by using Web Audio Api's decodeAudioData method.

var fileName = "pcmData";
var encodedName = "encoded.flac";

var args = [
  "--sample-rate=16",
  "--sign=signed",
  "--channels=1",
  "--bps=16",
  "--endian=big",
  fileName
];

var inData = {};
inData[fileName] = buffer;

var outData = {};
outData[encodedName] = {
  // Its MIME type
  'MIME': 'audio/flac'
};

worker.postMessage({
  command: 'encode',
  args: args,
  outData: outData,
  fileData: inData
});

Worker message gives me,

e.data.reply= err
e.data.values[0]= No such file or directory

[Rillke]

Hey @mertakdut

1. please use pcmData.raw as fileName
2. please use pcmData.flac as encodedName (most important bit here: either you catch the flac generated file or you specify the -o option to force an output file name)
3. buffer should be an Uint8Array.
4. --sample-rate=16000
5. Please submit an improvement to README.md (or to the code) so you or future visitors won't stumble over that issue.

var fr = new FileReader();
fr.onloadend = function() {

var fileName = "pcmData.raw";
var encodedName = "pcmData.flac";

var args = [
  "--sample-rate=16000",
  "--sign=signed",
  "--channels=1",
  "--bps=16",
  "--endian=little",
  fileName
];

var inData = {};
inData[fileName] = new Uint8Array(fr.result);

var outData = {};
outData[encodedName] = {
  // Its MIME type
  'MIME': 'audio/flac'
};
worker.onmessage = function() {console.log(arguments);}
worker.postMessage({
  command: 'encode',
  args: args,
  outData: outData,
  fileData: inData
});

};
fr.readAsArrayBuffer( document.getElementById('abc').files[0] );

[mertakdut]

Since I am getting float32Array from decodeAudioData output AudioBuffer.getChannelData, I need to convert the float32array to uInt8Array. When I use it with new Uint8Array(float32Array.buffer) the data is corrupted; only hearing sizzling sound.

Once I solve the problem, I'll provide the code for raw data encoding.

[Rillke]

When I use it with new Uint8Array(float32Array.buffer) the data is corrupted; only hearing sizzling sound.

Maybe there is a way to pass the float32-data directly to the flac command line executable. No time right now to look into this.

convert the float32array to uInt8Array

Looking for function floatTo16BitPCM?

https://github.com/mattdiamond/Recorderjs/blob/08e7abd99739be6946f19f6806ccb368138f2dd3/src/recorder.js#L142

The uInt8Array is for internal processing only getting offsets and file size calculations right etc. You shouldn't have to reduce the bit per sample (bps).

[Rillke]

Maybe upload your audio sample here so others can help?

[mertakdut]

Here is the raw data (I'll be breaking the copyrights if they decode & play the file :)). It can be tested with options (I suggest through Audacity File -> Import -> Raw Data...):

• Encoding: 32-bit float
• Byte Order: Little-endian
• Channels: 1 Channel (Mono)
• Sample Rate: 16000 Hz

I'll check the flac for direct float32 input.

[mertakdut]

Looking for function floatTo16BitPCM?

Tried this (and close alternatives) with "--sign=signed" argument.

var output = new Int8Array(input.length);

for (var i = 0; i < input.length; i++) {
  var s = Math.max(-1, Math.min(1, input[i]));
  output[i] = s < 0 ? s * 0x0080 : s * 0x007F;
}

return output;

Outputs:

pcmData.raw: ERROR: got partial sample

[mertakdut]

Okay. Now able to convert float32 to uint8 without corrupting IEEE 754 representation.

var output = new Uint8Array(float32Array.length);

for (var i = 0; i < float32Array.length; i++) {
  var tmp = Math.max(-1, Math.min(1, float32Array[i]));
  tmp = tmp < 0 ? (tmp * 0x8000) : (tmp * 0x7FFF);
  tmp = tmp / 256;
  output[i] = tmp + 128;
}

return output;

But issue11 gets in the way: Integer Overflow.

My tests with flac 1.3.2 win64, Uint8 encoded data with command flac --endian=little --sign=unsigned --channels=1 --bps=8 --sample-rate=16000 path/to/uint8comi have been successful.

Again the raw data can be tested by using Audacity's Import Raw Data feature. Encoding should be set to Unsigned 8-bit PCM.

[Rillke]

But issue11 gets in the way: Integer Overflow.

That is now fixed. Sorry for the long delay.

[mertakdut]

It's working now. Shall we put the pcm conversion example to readme for further information? Usage section is quite dull atm.

[pawohl]

@mertakdut Great :) Can you submit a pull request?

[mertakdut]

Done.