How to find the right frequencies
For simple inversion, it doesn't matter so much which exact frequency you use. It is best found by trial-and-error: Try various presets and listen which sounds best. There's a fair chance the audio will be somewhat intelligible with any setting.
If the audio is not intelligible after running it through deinvert with default settings it's likely not frequency inversion at all; it won't be worth the effort to search for the right frequency. Frequency inversion is very rare these days.
You can also use the above method for split-band scrambling. (But please note that deinvert does not support Variable Split-band.) If you don't specify the split point, deinvert will simply treat it as simple inversion. It won't sound perfect, but often it's intelligible. Try it! But don't get your hopes up - split-band inversion is very rare these days.
If you wan't better audio quality, then there's a trick for finding out the split point frequency. For this, you will need some software with a spectrogram display. For example, Audacity can show spectrograms.
Open the sound file in Audacity. By default, you will see the waveform (oscillogram), like below. Click on the little triangle in the top-left corner of the track and select "Spectrogram".
You will now see a frequency representation of the sound. Look for a frequency that is overall darker than all the other frequencies. This horizontal dark band, or 'dip', will quite likely correspond to the split-point frequency.
For example, in this file, the split-point frequency seems to be around 1.3 kHz, which corresponds to the option -s 1300.
The inversion carrier frequency (-f) can be guessed; try one of the preset frequencies listed in the readme. Or you could try using Audacity for that, too. Open the same menu that you saw previously. Select Spectrogram Settings... and set the maximum frequency to 6000. Now you will see a wider range of frequencies. A good guess for the inversion carrier is where the "noise" seems to end; in this case, close to 3200 Hz.
In conclusion:
- Often you don't have to provide a split-point frequency at all.
- If the audio is not understandable after inversion, probably it's scrambled with a different method, like VSB or time-splicing. Deinvert cannot descramble those.
- To guess the frequencies, look at the spectrum. You will see "dips" or quiet frequencies. These may correspond to the carrier frequencies.