Plangent can do amazing things--I've heard it used--but in addition to the high cost, it requires special playback equipment. It works on the assumption that the (high-frequency, usually between 60 and 100 kHz) bias oscillator in the original recorder had a steady frequency, at least on a short- to medium-term basis. To use the Plangent system you have to make a digital transfer from the original recording that includes the high-frequency bias. The system then reclocks the transferred recording so as to keep the bias signal at a steady frequency, which straightens out the audio signal almost as a side effect.
However, the remnant of such a high-frequency signal is always very weak at best, and is subject to a greater likelihood of further erasure simply because the wavelengths are so short, especially on slow-speed tape recordings. Plus nearly all analog tape recorders are designed to filter out the bias signal during playback. So Jamie Howarth (inventor of the process) has a set of decks of various kinds in which he's bypassed the filtering and/or tweaked the response so that it peaks in the bias frequency range. And of course he has to transfer the recordings at high sampling rates and/or do tricks such as playing back 15 ips tapes at 7-1/2 or 3-3/4 ips and then correcting the equalization.
In any case he then has to do a damned lot of numeric processing per minute or hour of recorded sound. It's definitely not a real-time process, even with plenty of computing power on hand.
This also tends to fix up the problem of recordings in which the overall speed shifts as the tape tension changes. That's a problem even on many professional recorders; the holdback and takeup tension would vary with the diameter of the tape pack on each reel, so typically the tape would move a little faster toward the beginning of a reel (or side of a cassette) and a little slower toward the end. That works out OK if you always play the same tape back in its entirety on the same deck. But if you cut and splice various takes together to make a master tape, there can be small pitch shifts at the splices.
Interestingly, rather than recovering horrible-sounding material, his system seems to get used most often for improving sound that wasn't at all bad to begin with. Some people seem to respond very strongly to this improvement; I have to admit that I don't hear it myself. Most likely I haven't caught on to the defect in the originals (sometimes a person needs training before something sounds wrong to them, and then they start hearing it everywhere).
I've also heard Capstan demonstrated. It works by analyzing the audio signals themselves, and requires human intervention in case it gets the wrong idea about a signal--left to its own devices, it can work like overzealous AutoTuning and strip the vibrato out of a violin, for example. But if you have (IIRC) $4,000 burning a hole in your pocket, you can buy the software and use it to your heart's content, and I'm sure it will come down in price eventually, especially if some competition emerges.
--best regards