Landing page with summary, block diagrams, screenshots, and demo video:
https://github.com/OpenV2K

OpenV2K is an open-source Python application that converts arbitrary audio — including both natural speech (via live microphone), and diphone synthesis (via text-to-speech engine) — into pulse-modulated radio-frequency (RF) waveform. The signal chain can be used to transmit via a software-defined radio (SDR), or to save the raw IQ samples to disk, and generate spectrograms automatically. Integrates toggled DSP filters, into a PyQt5 GUI. Intended to be usable by technicians.

The project's signal chain is based on microwave auditory effect literature, most notably Sharp and Grove's demonstration published in 1975, that pulse-modulated microwave energy could convey intelligible words, building on Frey's work. The app currently integrates Sharp's modulation technique: detects every zero-crossing of the audio waveform, as depicted in the GitHub profile thumbnail. Home users get to see, what made the US Army's "V2K", and US Navy's "MEDUSA", function, in real-time.

This prototype was built on the shoulders of giants. Thank you to SME biophysics researchers, that publicly released their work, 5 to 50 years ago; the FOSS software package developers; an the open-source SDR hardware platform developers. I'm an IT systems engineer, with several years of professional programming experience, and I've been wearing a project manager hat, with Claude.

Summary of associated research from James C Lin, an IEEE life fellow, and biophysics/EE professor, who wrote the 2021 book Auditory Effects of Microwave Radiation. From chapter 10.3, page 326:

Over the years, some have expressed concerns about the capability of communicating directly with humans by pulsed microwaves. In fact, earlier while studying microwave-induced auditory effects in human subjects, it was noticed that in addition to zip, click, and knock sounds from exposure to single pulses of microwave radiation, short trains of rectangular microwave pulses are heard as chirps with tones corresponding to the pulse repetition frequencies [Guy et al., 1973, 1975]. It was also found that when the pulse generator was keyed manually such that each closing and opening of a push-button switch resulted in emitting a short rectangular pulse of microwave energy-transmitted digital codes (Morse code) were received and accurately interpreted by the targeted subject [Lin, 1978].

Also, soon after a system was setup for exploring the microwave auditory effect at the Walter Reed Army Institute of Research (WRAIR) [Sharp, et al., 1974], two of the principal investigators were able to demonstrate direct communication of simple speech via appropriate modulation of microwave energy [Justesen, 1975]. They tape-recorded each of the single-syllable words for digits between one and ten. The speech waveforms of each word were then converted to digital signals in such a fashion that each time an analog speech wave crossed the zero reference in the negative direction, a short pulse of microwave energy was emitted from the transmitter. By subjecting themselves to the exposure of "speech modulated" microwave energy, the investigators reported the ability to hear, identify, and distinguish the words transmitted.

These historical experiments are included to show that this has been accomplished before. There are safer methods to demonstrate this today, that do not involve anyone's head.

Download from GitHub, Public Domain:
https://github.com/OpenV2K/Sandbox