User documentation

Introduction

All the functionality of the program is available from the main window. In this window you can configure parameters for spectrogram generation or synthesis and supply the data.

Progress of long operations is indicated on the right side of the window. The X button can be used to interrupt an operation that is taking too long.

Spectrogram analysis

To turn a sound to a spectrogram, select the sound file in the upper right part of the window. Many sound files are stereo, which will appear as two channels you can choose from. The "Length" and "Samplerate" indicators are purely informative.

Depending on the purpose of the spectrogram and the nature of the supplied audio data, different parameters are optimal. The meaning of the main parameters is explained below. If you hover your mouse over a parameter dial, an explanaition appears as a tooltip.

• Frequency scale: This setting determines the type of the frequency (vertical) axis. Human hearing is logarithmic in nature. For music and other audio that contains a high range of frequencies, logarithmic frequency scale is a good choice. For speech or artificial sounds, linear scale can also be used with good results.
• Intensity scale This affects the mapping of sound intensity to pixel brightness. The logarithmic setting is better for sounds with high range of loundess where a linear setting would make the spectrogram too dark.
• Base frequency For a logarithmic frequency spectrogram, the first band (the row of the spectrogram) will be centered at this frequency. For a linear frequency spectrogram, the first band starts at the this frequency.
• Maximum frequency Sets the top frequency displayed in the spectrogram. For speech, value of about 8000 Hz can be sufficient.
• Pixels per second Determines the time resolution of the spectrogram. The larger the value, the wider the spectrogram. For synthesis, 100 or above is recommended. For viewing, 50 can be sufficient and make the spectrogram more "compact" with long sound samples.
• Brightness correction Some spectrograms can be very dark even with logarithmic intensity scale. Using the square root brightness correction will make the spectrogram easier to read, but may affect synthesis quality.
• Bandwidth Each horizontal band of the spectrogram will be as wide as set here. Lower value means more detail in the frequency domain, but less detail in the time domain.
• Window function: Window function is applied to the frequency intervals of the given bandwidth to lessen artifacts on their edges. The Hann window is a good general choice.
• Overlap: Larger overlap gives more detail in the frequency domain and makes the spectrogram taller. If no window function is used, it can be set to zero, otherwise setting at least 60% overlap is recommended.
• Palette: Shows the colors in which the spectrogram will be drawn. You can supply your own palette from an image, in that case the first row of pixels of the image is used. For synthesis, the colors in the palette shouldn't repeat to make the intensity -> color mapping unambiguous.

Once you are happy with the parameters, click the "Make spectrogram" button. A preview will appear and you can save the resulting image.

Spectrogram synthesis

To turn a spectrogram back into sound, first select the spectrogram image in the lower right.

The parameters and palette of the spectrogram should be set to the same values as were used for its generation. If the spectrogram was generated by this program, the parameters will be loaded automatically from metadata saved in the image. You can work with spectrograms from different sources too, in which case you need to know or guess the parameters with which they have been created.

Two modes of synthesis are provided:

• Sine synthesis is fast and produces decent results.
• Noise synthesis is slower but may give better results for "busy" spectrograms.

When the parameters are set, you can press "Make sound" to synthesize the chosen spectrogram. After it's finished, you can save the resulting sound file.

Supported file formats

The program supports most commonly used sound file formats like mp3, wav, flac and ogg. For the last two the build has to be linked with the libsndfile library version 1.0.18 or higher with libflac and libogg plugins built-in. See http://www.mega-nerd.com/libsndfile/#Features for a full list of sound file formats supported via libsndfile. MP3 is supported via libmad.

Certain MP3 files with variable bitrate can display the wrong length in the GUI. They can still be used to generate spectrograms without problems.

Most commonly used image formats are supported, for example png, bmp, tiff, xpm, jpg (read only), gif (read only) and others. See http://doc.trolltech.com/4.6/qimage.html#reading-and-writing-image-files for a full list of supported image formats.