Yeah, I've been chewing on that stuff for a while. First thing: keep everything lossless—WAV or FLAC—because any compression will eat the clues. Then isolate the high‑frequency tail, run a fine‑grained spectrogram, and look for tiny periodic modulations that repeat. If you see a regular ripple you can isolate that band with a band‑pass filter and run an FFT to decode the hidden waveform; just be careful not to blur the edges, otherwise you’ll kill the signal and the fidelity.

I don’t have ready‑made demos on hand, but you can whip up your own in minutes. Grab a clean recording of a song in WAV or FLAC, then use a tool like Audacity to embed a low‑amplitude sine wave in the top 15‑20 kHz band. Export it again as lossless, and you’ve got a test track. If you want something from the wild, search for “audio steganography samples” on GitHub or the Internet Archive—there are a few free packs that use the same techniques. Once you’ve got a sample, start hunting with the spectrogram trick I mentioned before. If you’re feeling brave, just dive in and experiment with your own recordings—no one can do better than a perfect copy of your own ears.

Just make sure the sample rate is at least 192 kHz so 18 kHz sits well inside the Nyquist band, use a Hamming window on the FFT so you don’t get spectral leakage, keep the hidden sine at a level that’s below the noise floor but not so low it gets lost in quantisation, and double‑check the envelope for any clicks or spikes that would ruin the pristine mix. Also run a quick bit‑depth check—if you’re working in 24‑bit, keep the hidden signal in the top few dB; if it creeps into the bottom 10 bits you’ll start hearing quantisation noise. That’s all the nitty‑gritty you’ll need to keep the fidelity intact while hunting those secret bits.

Got it, no worries. If you ever feel like giving it another go or just want to chat about the next tweak, hit me up.