Photosounder: editing a sonogram

I publish this topic at the request of HammeremmaH, mired in the minus. I hope that with your help he will be able to publish his next article on his own.

Have you ever thought about how great it would be to work with sound in a graphic editor?
- Here is the violin. Where is my healing-brush, now let's cover it up so that nobody even noticed.
- But the vocals, see the "drunk" stripes? Come here your eraser - we make a phonogram.
This is how you can describe the process of editing a sonogram in Photoshop.

Photosounder is a software that allows you to analyze a sound fragment, to build on its basis a “canvas” of frequency-time-loudness dependencies. Moreover, this “canvas” can be exported as an image, edited, and imported back into the program, where changes in the image will be converted to sound. Sounds fantastic?
')

Theory. How it works?

The FFT algorithm (FFT - Fast Fourier Transform) is used to analyze and construct a sonogram. As a result of the FFT-algorithm, the sound is decomposed into all its component frequencies. Information on each frequency is analyzed at each time interval and as a result an image is built, where time is plotted along the X axis, frequencies are plotted along the Y axis, and loudness at the intersection of coordinates is expressed by pixel brightness. This is how it looks in practice:



The program allows you to export the sonogram as a BMP image, as well as import the image of the sonogram back. With the help of our favorite graphic editor, we can make any changes to the sonogram, as a result of which the frequency content of the future sound will change.

Here is a small demo:


And here is how this demo was done:

What does this give us?

With a certain level of skill and zeal, you can make excellent soundtracks without losing the quality of the original. You can also use the effects of a graphic editor to get a completely new sound. But what is even more interesting, using this technology, you can slow down any sound any number of times without loss of quality and without the effect of granulation (which is used in almost all time-stretching algorithms). Some examples on offsite .