I've just added a new program to Marsyas called sound2png. It lets you make a picture of a an audio file in .mp3, .wav, .au or .aiff format:
When generating a spectrogram, you can set both the window size and hop size that are used in calculating the FFT. The window size that you give is then used as the amount of data that the FFT is given, which means that the number of bins for the FFT will be half of the window size. Each bin of the FFT will be drawn in one pixel vertically, so if you use a window size of 512, the resulting PNG will be 256 pixels high.
The hop size for the spectrogram tells the program how much to overlap each FFT by. The width of the output PNG will thus depend on the length of the audio file and the hop size, with smaller hop sizes giving longer PNG images.
Below is shown an example of using sound2png to generate a spectrogram of an orca call. We use a window size of 1024 and a hop size of 1024. The maximum frequency is set to 8000Hz. A gain of 1.5 is used to make the spectrogram darker:
sound2png A30.wav -ws 1024 -hs 1024 -mf 8000 -g 1.5 out.png
You can also you sound2png to generate pictures of the waveform of an audio file. For this, you use the -w option. An example of this is shown below:
sound2png tiny.wav -ws 1 -w out.png
When generating pictures of waveforms, you can specify a window size. sound2png takes a chunk of data that is window size samples in length and calculates the maximum and minimum of this window. It then draws a bar from the minimum to the maximum value for each window. An example of this is shown below:
sound2png small.wav -ws 100 -w out.png
