You need to hear the sound of this filter to actually understand its quality. I took the liberty of making a sound demo that captures the performance of the pumpkin filter.
http://www.iface.ne.jp/~ken/pumpkinfilterdemo.mp3
In the beginning half, you can hear how the pumpkin filter has rebuilt the low end, mids, and high end to sound perfectly similar to the original signal. I suppse Celmo has used some sort of STFT reverse synthesis mechanism.. still, pumpkin filter can rebuild the signal so much alike the orignal signal, that perhaps there are some hidden secondary resynthesis algorythms. I'm guessing there is some form of simple granular synthesis going on... or some sort of sample manipulation, as the reconstructed sound is sample accurate when compared to the original signal.
In the second half, the pumpkin filter has colorized the signal to sound like silence. As you can hear, Celmo's "silence" has a certain warm, phat, vintage feel to it that is reminiscent of all his other devices. I tried comparing pumpkin filter's "silence" to silence from Cooledit, Soundforge, Wavelab, and even ModIII, but Pumpkin filter's silence had most character, being totally dead, dark, and so.. "not there".
All in all, this device is a completed, polished product for the most professional musician. 20 years ago, this feat may have seemed impossible to achieve with the slow processors and weak DSPs of the time. Years have gone by, and now, finally, we can have perfectly modeled, physically accurate, totally custimizable silence at our fingertips.
Got a bad mix? Run it through pumpkin filter and shut it off.