Silk SDK

This is node-gyp version of Silk SDK library (see https://developer.skype.com/silk for details.)

Installation

To install this package, please run:

npm install silk-sdk

or

yarn add silk-sdk

If you wish to use cli tool, try install globally using:

npm install silk-sdk --global

or

yarn global add silk-sdk

Command Usage

decode

To decode silk format audio, run:

silk-sdk decode [input] [output]

where input is the name of input file and output is the name of output file. You can pass further options after, availables are shown below:

• -q, --quiet Whether there is any output into console (default: false)
• --lossProb [lossProb] Simulated packet loss percentage (0-100) (default: 0)
• --fsHz [fs] Sampling rate of output signal in Hz (default: 24000)

encode

To encode audio to silk format, run:

silk-sdk encode [input] [output]

where input is the name of input file and output is the name of output file. You can pass further options after, availables are shown below:

• -q, --quiet Whether there is any output into console (default: false)
• --fsHz [fs] API sampling rate in Hz (default: 24000)
• --fxMaxInternal [fxMaxInternal] Maximum internal sampling rate in Hz (default: 0)
• --packetLength [packageLength] Packet interval in ms (default: 20)
• --rate Target bitrate (default: 25000)
• --loss Uplink loss estimate, in percent (0-100) (default: 0)
• --complexity Set complexity, 0: low, 1: medium, 2: high (default: 2)
• --inbandFEC Enable inband FEC usage (default: false)
• --dtx Enable DTX (default: false)
• --tencent Add Tencent (Wechat, QQ) header in exported file (default: false)
• --tencentAmr Add Tencent AMR header in exported file (default: false)

compare

To compare to silk format audio files, run:

silk-sdk compare [ref] [test]

where ref is the name of first file and test is the name of second file. You can pass further options after, availables are shown below:

• -q, --quiet Whether there is any output into console (default: false)
• --diff Only determine bit-exactness (default: false)
• --fs Sampling rate in Hz, max is 48000 (default: 24000)

API Usage

decode

This API will decode silk format audio. It's return value and behavior will change depends on the parameters provided.

Stream.Transform decode(options)

Parameter	Type	Required	Description
options	object	false	Optional options, see below for details

This API will return a transform stream, which can be used to consume input stream and generate decode stream.

Exmaple usage:

fs.createReadStream('input-path-here')
  .pipe(silk.decode({ quiet: true }))
  .pipe(fs.createWriteStream('output-path-here'));

buffer decode(input, options)

Parameter	Type	Required	Description
input	string/buffer	true	Bitstream input to decoder
options	object	false	Optional options, see below for details

For first parameter input, you can pass in either string or buffer. If string is provided, it will be used as path of file; if buffer is provided, it should be the buffer of input file.

This API will return a decoded buffer that can be either saved in file or used by other APIs.

Example usage:

fs.writeFileSync(silk.decode('input-path-here', { quiet: true }));

void decode(input, output, options)

Parameter	Type	Required	Description
input	string/buffer	true	Bitstream input to decoder
output	string	true	Speech output (file path) from decoder
options	object	false	Optional options, see below for details

For first parameter input, you can pass in either string or buffer. If string is provided, it will be used as path of file; if buffer is provided, it should be the buffer of input file.

No return for this API, decode result will be saved to output path directly.

Example usage:

silk.decode('./input.silk', './output.pcm', { quiet: true });

Options

Attribute	Type	Default	Description
quiet	boolean	true	Whether there is any output into console
lossProb	float	0	Simulated packet loss percentage (0-100)
fsHz	int	24000	Sampling rate of output signal in Hz

encode

This API will encode speech input to silk format. It's return value and behavior will change depends on the parameters provided.

Stream.Transform encode(options)

Parameter	Type	Required	Description
options	object	false	Optional options, see below for details

This API will return a transform stream, which can be used to consume input stream and generate decode stream.

Exmaple usage:

fs.createReadStream('input-path-here')
  .pipe(silk.encode({ quiet: true }))
  .pipe(fs.createWriteStream('output-path-here'));

buffer encode(input, options)

Parameter	Type	Required	Description
input	string/buffer	true	Speech input to encoder
options	object	false	Optional options, see below for details

For first parameter input, you can pass in either string or buffer. If string is provided, it will be used as path of file; if buffer is provided, it should be the buffer of input file.

This API will return a encoded buffer that can be either saved in file or used by other APIs.

Example usage:

fs.writeFileSync(silk.encode('input-path-here', { quiet: true }));

void decode(input, output, options)

Parameter	Type	Required	Description
input	string/buffer	true	Speech input to encoder
output	string	true	Bitstream output from encoder
options	object	false	Optional options, see below for details

For first parameter input, you can pass in either string or buffer. If string is provided, it will be used as path of file; if buffer is provided, it should be the buffer of input file.

No return for this API, encode result will be saved to output path directly.

Example usage:

silk.encode('./input.silk', './output.pcm', { quiet: true });

Options

Attribute	Type	Default	Description
quiet	boolean	true	Whether there is any output into console
fsHz	int	24000	API sampling rate in Hz
fxMaxInternal	int	0	Maximum internal sampling rate in Hz
packetLength	int	20	Packet interval in ms
rate	int	25000	Target bitrate
loss	int	0	Uplink loss estimate, in percent (0-100)
complexity	int	2	Set complexity, 0: low, 1: medium, 2: high
inbandFEC	boolean	false	Enable inband FEC usage
dtx	boolean	false	Enable DTX
tencent	boolean	false	Add Tencent (Wechat, QQ) header in exported file
tencentAmr	boolean	false	Add Tencent AMR header in exported file

compare

bool compare(input, output, stream)

Compare two audio files.

Parameter	Type	Required	Description
inputA	string/buffer	true	Reference file
inputB	string/buffer	true	File to be tested, should be of same length as inputA (pcm)
options	object	false	Optional options, see below for details

For both inputA and inputB, you can pass in either string or buffer. If string is provided, it will be used as path of file; if buffer is provided, it should be the buffer of input file.

Below are possible options for compare.

Attribute	Type	Default	Description
quiet	boolean	true	Whether there is any output into console
diff	boolean	false	Only determine bit-exactness
fs	int	24000	Sampling rate in Hz, max is 48000

Example usage:

const silk = require('silk-sdk');
silk.compare('./ref.pcm', './test.pcm', { quiet: true });

License

This wrap of original Silk SDK code is under MIT License. For Silk SDK itself, please check their license for details.