docs: How to get value of frequencies?

[BigGitWorld]

Hello.
I want to plot a Line chart from frequency domain of a wav file (x-axis: Frequencies, y-axis: decibel values).
How can I get value of frequencies?

[alnitak]

Hi,
you can only get FFT and wave data while the audio is playing. See getFft(), getWave(), getAudioTexture() in the doc.
They return a convenient Pointer<Float> for performance instead of copying it ie in a Float32List every time.

Getting a whole audio frequency domain at once is not yet planned

[BigGitWorld]

Hi, you can only get FFT and wave data while the audio is playing. See getFft(), getWave(), getAudioTexture() in the doc. They return a convenient Pointer<Float> for performance instead of copying it ie in a Float32List every time.

Getting a whole audio frequency domain at once is not yet planned

Thanks for your answer.

1. But I didn't understand how can I get value of frequencies to Hz from FFT data. Do you mean ffi.Pointer<ffi.Float> audioData represents amplitude of each sine wave or frequency of sine wave? I want to have both of them (a Map with key=frequency and value=amplitude of that sine waves that create our signal). How can I find it?
   could you give me source code of this example where x-axis shows frequencies and y-axis shows amplitude of them to decibel?

2. In fact, I have a reference wav file and want to compare other wav files with it to check if they are similar or not.
   can you suggest me a solution please?

[alnitak]

I notice now that I didn't exposed getAudioTexture(), it's not available on Flutter side. So you will need to use getAudioTexture2D().

Using getAudioTexture2D() you will get back a matrix with 256 rows of 512 floats. You could consider only the 1st row as it represents the latest data. So this first row is composed by:

• The first 256 floats represent the FFT frequencies data in the 0.0~1.0 range > 0.0.
• The other 256 floats represent the wave data in -1.0~1.0 range (amplitude).
  So, the first 256 floats in this array, I think are the values you are looking for:

final ffi.Pointer<ffi.Float> audioData = SoLoud().getAudioTexture2D(playerData);

You can access these floats with double n = audioData[i]. The i index is the frequency in Hz and its value is the amplitude (volume) of that frequency.
Since the engine is set to 44100 samples/s and the amount of data given to compute FFT is 1024, below a function to know which frequency is represented in the array index i.

double Index2Freq(int i, double samples, int nFFT) {
  return  (samples / nFFT / 2.0) * i;
}

and you can use it with something like this:

void getFreqs() {
  for (final int i=0; i<256; i++) print('index: $i  is  ${Index2Freq(i, 44100, 1024)} Hz');
}

Here you can see the widget used in the Visualizer example to show the FFT data, but the correct image is this. The one you linked is made using Audacity app and it's a bit different with the one generated by this plugin and I don't know yet why 😄

I didn't tested and I don't know much about audio processing, but hope this helps!

2. I think this could be more complicated than it might seems. To compare audio you should use fingerprints, basically you should generate these fingerprint of samples and then compare each other. There are some libraries around by I never tried.

I move this issue to Discussion, it maybe be useful to others!

[BigGitWorld]

1. why the The first 256 floats that represents FFT frequencies data is in range of (0.0~1.0)?
2. why the second 256 floats that represents wave data is in range of (-1.0~1.0)?

[BigGitWorld]

No reply?

[alnitak]

Oh! Sorry, I missed it!
By default, float values are used for faster interaction with audio data. It's possible, to modify the plugin code (not exposed to Flutter) and use integers, but I have chosen to stick with float numbers.

Wave data is in the -1~1 range because I choose to use Soloud::CLIP_ROUNDOFF. Here some Soloud docs about clipping.
FFT data has values >0 (I have edited the previous post). This means that values can be greater than 1.0 and it is due to the windowing method used. Here I used Blackman windowing which is used also by ShaderToy, even if I'm not fully satisfied.