I would like to transmit 24 bit 96 kHz audio
so, 2.3 Mbit/s (that's ca 63.6 dBbit/s)
(maybe up to 30 metres)
Say, for an ca 900 MHz that means about 60 dB path loss, so at a good-will of 20 dBm transmit power (good luck getting that throug regulation), -40 dBm receive power, no matter the bandwidth \$B\$.
So, that gives us -40 dBm for all our 63.6 dBbit/s, and that means -103.6 dBmJ per bit = 10⁻¹³ J per bit of energy, or 0.0000000000001 J. That's not exactly much.
Thermal noise at room temperature (not: body temperature; not: stage temperature in broad light) is -174 dBm/Hz, leading to a noise density of \$N_0=-174\,\text{dBmJ}\$. Oh, we're assuming totally noise-free active parts in the radio receivers, by the way. Totally not happening, but what's a factor of 10 among friends?
You hence got an \$\frac{E_b}{N_0}\$ of ca 70 dB – not too shabby, actually.
So, that gives you a channel capacity of
$$C=B\log_2\left(1+\frac{E_b}{N_0}\right) = B 7\log_2(10) \approx 23B$$
Hence, with a \$C_{min}\$ of 2.3 Mbit/s,
$$23B \ge C_{min} = 2.3\cdot 10^6 \frac{\text{bit}}{\text{s}}$$
follows a minimum bandwidth of 100 kHz.
at a latency that is less then 5 milliseconds
So, within the time it takes to take 480 samples, process them, channel code them, modulate, transmit, receive, demodulate, channel decode and reproduce them.
That means we can't work with long channel codes, and hence, we can't achieve capacity. We'll need some more bandwidth – say, 250 kHz, at the very least. Get looking for low-complexity low-latency channel codes; you'll often find pretty "ancient" codes being used here that are very far from optimal, but pretty limited in memory.
Now, you "only" need to implement a few more things: framing, scrambling, frequency and phase and amplitude synchronization, timing recovery, symbol decision, interleaving, analog to digital and digital to analog for the RF signals, and the whole audio system :)
On the contrary, you cite this, where someone simply takes an SPDIF data stream and sends it over 6 MHz of analog NTSC transceiver and calls it "lossless audio":

You want to build this so-called "$12 wireless lossless digital audio", to compete with, according to your whole question, this:

- Lossless is a lie; this is an analog transmission link and it most definitely has loss of bits, and hence fidelity (you might not hear it, because at the crass oversampling of 96 kHz, which is totally unnecessary for human hearing, fight me, I have the math, the chance of modifying a significant bit often enough in a row are small enough)
- "Something like these 6.35 mm plug-sized adapters that work off a battery for the duration of a concert" vs "thing that is a 6 MHz NTSC transmitter": I trust you'll notice the difference in size and in power consumption yourself.
- The NTSC thing works over "7 feet". I had to convert that to meters, it's a little more than 2m. You want to work over 30m, which means you get about 200 times less power at the receiver.
Seriously, this hacked up 12 $ something of questionable sense has no comparable features to the professional 190$ stage equipment you're trying to build.