The general concept is the following. There are six key 'frames' which can be edited manually. It is possible to generate intermediate sounds by interpolation between the frames.
The overall duration of the generated sound is limited only by available memory and your patience (the calculation might appear time consuming): please note the Steps control which determines how many intermediate sound frames would be generated between the 6 given reference sound frames. Another point is that every frame might be as long as 10 seconds - that gives a minute if number of steps is 1, two minutes if it's 2, and so on.
Each frame might have different duration.
Each frame is built from three graphs: amplitude, frequency and distortion.
Each graph has a context menu available by right-mouse-button clicking.
There are options of smoothing the graph, and so on.
The left mouse button is used to draw the graphs.
There is 
button which allows to copy one frame into another.
The Random button generates random amplitude and frequency for the selected
frame.
There is the Preview button which allows to hear the sound of every separate frame and see its spectral map. When all the frames are ready either Play or Save button would generate the complete piece which will be played or saved afterwards.
On the roadmap:
- The capability to import another data sources - like *.mid, *.wav and *.bmp (by the latter I mean graphs like EEG which could be scanned and used to produce somewhat 'natural').
- The capability to 'snap' the manually drawn graphs to fixed rhythm patterns.
- The capability to 'harmonize' sounds by adjustment instant frequencies to musical scales.