The Digital Phase Vocoder - Sound Mutation (Mad as Hell Speech)

"Mad as Hell" speech by Peter Finch as Howard Beale from "Network" (1976) mutated with Wolfgang Amadeus Mozart's Requiem (KV.626) "Lacrimosa" utilizing the Digital Phase Vocoder.

Category:

Science & Technology

Tags:

• Phase
• Channel
• Vocoder
• Sound
• Mutation
• Network
• Peter
• Finch
• Howard
• Beale
• Mad
• as
• Hell
• Wolfgang
• Amadeus
• Mozart
• Lacrimosa

[tizwah]

Never heard of that book, sorry. I'm familiar with the classic text books on speech processing though, such as Rabiner/Schafer et al. Your response, while true, doesn't change the fact that the channel vocoder follows a different routing and application paradigm. Even though you can think of the pvoc as being a superset of both types of vocoder they are different in many ways, historically and wrt implementation/application (see my previous comment for details).

[johgruenwald]

@tizwah Thanks for the response, but I fear you didn't get my point: "The channel vocoder yields the same effect, but from a source-filter processing approach" (see above) is exactly what you keep trying to tell me. I only stated that two different approaches can yield the same result, and the fact because either is more common doesn't allow one to propose it to be the only/better one. Anyway, I got your point and I hope you got mine :) For further reference I highly recommend the DAFx book.

[neogeo53]

that was such a great video i loved the vocoder put on his voice

[tizwah]

@johgruenwald Sure sounds good and thanks for the hint. Btw. there is a great tutorial on pitch shifting using the Fourier transform on Stephen Bernsees DSP Dimension page (google for "Bernsee code") to find it. Highly recommended and comes with C source code.

[johgruenwald]

@tizwah (3) consisting of phase of the one and amplitude of the other input source. The channel vocoder yields the same effect, but from a source-filter processing approach, which was not the topic of my thesis. Please note that I am speaking from a theoretical signal processing point of view and a resulting MATLAB implementation. Realizing this effect in hardware for musical application is a complete different topic.

[johgruenwald]

@tizwah (2) "mutation between sounds" (which we are speaking about) are - besides the "classic" phase vocoder effects pitch shifting / time stretching - thus phase vocoder effects as well. Concerning the mutation effect, it's not a considerable alternation but very straight forward simplification: there is no pitch shifting (thus there is no complex stuff in the analysis / synthesis stage besides windowing and performing FFT/IFFT), and in the processing stage, i just set up the new STFT frame

[johgruenwald]

@tizwah (1) I see, you have theoretical knowledge here. Referring to the book "DAFx" (edited by Udo Zölzer), which is a well established reference of digital audio effects in terms of digital signal processing, the phase vocoder is specified consisting of an analysis stage and synthesis stage (windowed FFT/IFFT, instantaneous phase computation etc.), surrounding the time-frequency processing stage which applies any DSP algorithm on STFT frames. Effects such as "robotization", "whisperization",

[tizwah]

@johgruenwald Yes but the whole point of a phase vocoder is access to temporal evolution of sound and have the ability to reconstruct sound by including instantaneous phase in the process. It does not use a carrier/modulator relationship at all. A channel vocoder is actually quite different. You can use a phase vocoder to build it, but you would have to alter it considerably (group bins, add a second input, discard phase) so it's really no longer the same process...

[johgruenwald]

@tizwah well, as phase vocoders can implement channel vocoder effects as well i don't see a mistake here. i should have stated this detail more clearly in the description, you are right. but as the term "phase vocoder" was the overall title of my thesis, it was my first intuition to call it this way.

[tizwah]

This isn't a phase vocoder, it's a channel vocoder like the Prosoniq OrangeVocoder