I will die on this hill: FL users are either the newbiest producers, clipping their masters at +5dB, or they are borderline NASA scientists. There is no in between😂
@@disivni any csv file with one column can be a wav file. any wav file can be a csv with one column. there IS common ground, someone just needs to put it to use.
This is a geometric explanation of why taking the Fourier transform twice gives you the time domain back, only reversed. It's great to see another way of looking at it.
@@vinceb8123 This is only really relevant if you like messing with frequencies. Fourier's transform is the best way to map time-domain bins into sine-space for further analysis
you have no goddamn idea how ecstatic i am to learn that something i randomly conceptualized a long time ago is now actually possible ive found myself making more IDM, glitch, noise, and botanica music and this is such an insane gamechanger for me
sincerely congrats for figuring this stuff out, simply brilliant, maybe it seems to you that its goofy and gimmicky, or not very useful right now, but youre tapping in unexplored territory here and this could be the beginning of something bigger! ❤
the classical formality really is silly; i’m actually hoping to develop a syllabus to teach introductory applied signal processing with music/audio. A lot of people seem to benefit much more from seeing something applied and trying it out before learning all the math behind it.
you can create some crazy distortion if you rotate it once, add some kind of delay, and then rotate it back. Could probably do some crazy sound design with this
That's actually genius! What other effects could be useful swapping out the frequency domain for the time domain? Reverb to make stuff into weird white noise, I guess? Would EQ be like a particular volume automation curve over time? I'm not even sure what a flanger or phaser would I'm not even sure what a flanger or phaser would sound like, messing with all the frequencies in a sweep? Really cool idea to use the rotation like that!
@@NZsaltz perhaps this can be used to create distortion/saturation without aliasing? You'd just need a way to apply conventional 'distortion' in the time domain somehow... Wouldn't really be useful in a realtime setting, but probably useful in some way.
@@NZsaltz a flanger is actually caused using delay. Rotating it would sound pretty similar. A phaser, on the other hand, would depend on how many bands it had, but would be a weird comb-filtered delay I believe.
@@ekut1922 i started doing it. right now im looking into ways of making spectrogram. started with fft at first, but switched to Constant Q bcs its logarithmic (maybe ill make it possible to switch between them in the final version)
This is so cool, I was so surprised that so little was lost after four rotations?! You could use this to hide a sneaky easter egg or ARG clue in a song :3
well it is a visual thing, youre seeing it do the thing in the spectrum lol if its up to everyone to figure out uses for it yesterday i thought of "oh you can hide a vocal watermark in a beat with this without it sounding too ass if you layer it with other stuff and speak very clearly"
I've been wondering about the possibility of flipping the frequency response of an audio source upside down for ages now but never had the know-how to do anything like this, thank you so much for satisfying that particular curiosity
dude i cant get over how unique this is, it's such an otherworldly sound design idea, that's awesome!! imagine if someone made a hybrid of this spectral rotation thing with granular synthesis... like a plugin which would take each grain and rotate them to some angle. maybe with some damping on high freqs to make them more controlled, but i honestly think this idea could spark a whole new plethora of ways to do sound design...
Thanks for showing us how to do this. I found out if you bitcrush a spectrogram rotated 90 degrees and rotate it back it creates a repeated granulized sound?? Im super confused why but it sounds cool.
If you think about it, what any distortion does is just generate harmonics, in other words they generate information that is offset in the Y-Axis of a spectrogram. But if you rotate it, that information is now offset in the X axis - meaning you've generated information that is offset in time. If I'm understanding it correctly, this means a delay plugin alongside this method could be used to generate harmonics in a novel way.
this is so timely, I was just thinking how I wanted to do some spectrum range-shifting work (shifting, rotating, squashing, etc) the other day and have never done it before, now I have a few points to jump off at, cheers.
rotating something 90 degrees isn't a super lossy process so you aren't altering the original work THAT much, its a reversible process so its basically a novel way to store the signal weirdly
@@idiotinium if you rotated it 90deg, 180deg, or 270deg, no they can't i was assuming they were just rotating it 90deg or something and not a full 360deg
never seen this done in realtime before, every other thing like this exports the spectrum as an image, opening it in an editor, flipping it, then importing it back in. absolutley wild
This is like a brand new sentence but with lore behind it. I imagine cutting away the sweeping spund at the beginning and end contributes to the quality loss
Very much like the refractions of light seen in the glass of a window, a phone. Even in a fucking fl studio, you can see the fundamentals of the universe. Man, that's fucking awesome.
The most insane thing I've learned on my own while playing with GNURadio Companion flowgraphs. FFT itself does EXACTLY this. In one shot. If you pass the studied signal through the rectangular-windowed FFT (I mean, a really large one, by the width of the whole audio length ceiled to 2^n instead of the default 1024 samples) and then interpret the result as the ordinary signal, you will get the spectrum rotated 90 degrees. Both the source and the result must be the I/Q signal with two channels (as we are operating with complex float32 values). ...Basically, Fourier transform just rotates the time-frequency domain 90 degrees counter-clockwise. Replaces peaks with sines, and sines with peaks. The idea is THAT simple. ...So, if you apply FFT 4 times in a row, the result will be the original signal as we just rotated the frequency domain 360 degrees.
To anyone that wants an easier way of doing this, I recommend photosounder, with which you can edit audio as if it were a picture, which includes layering, rotating, changing contrast etc.
do not do this if you're trying to have a good outcome lol, you will have no phase spectrum and quantize your amplitude bins to the bit depth of the image you're saving. additionally think about the pixel interpolation by reading the image and the window function used to generate the spectrum image, also window overlapping etc. that process is supremely lossy
That is brilliant! I imagine you could do something similar if you manipulate the spectrum with an image editor and a tools Photosounder or Coagula, but it's amazing to do it strictly in the audio domain, so to speak! A lot of potential for interesting sound design, and of course Easter Eggs :)
keep in mind the phase spectrum is a very critical part of any signal's fourier decomposition, most of those tools only export magnitude/amplitude spectra, also with no window overlap or anything (some even distort it such that it no longer has a linear frequency axis). plus the partials' amplitude will be quantized to the bit-depth of the image further losing information keeping it in audio form retains phases and stuff cus its the pure actual math on the signal
I can think of exactly 0 use cases for this other than maybe encrypting something like someone said in a comment below, but it sounds sick. How did you even find out how to do this? I have no understanding of why what you did achieved the result. Edit: nvm, I watched your previous video about this, and it makes a lot of sense, this is insane lol.
i saw a recent matt parker (standupmaths) video explaining how a rotation is equivalent to 3 skews if you use specific ratios for the skews, and i also noticed that convolving with a sine sweep was one way to extremely offset phases, and a frequency shifter skews vertically when automated linearly so yeah also it would be a cool way to hide a secret in some puzzle thing for sure!
This would be interesting to do at a sample rate of say 20khz or 24khz so when it's inverted and backwards after 180deg rotation the majority of the audio is still within the audiable range.
This is super interesting. I have a few ideas with this: Harmor can actually process images to generate sound. Also, Melda’s MCabinet is a Convolver that only takes tonal information and not phase/time So it could provide a different result
maybe, i wouldn't count on it since it's linear instead of how notes are a geometric progression instead of arithmetic notes have an exponential scale while this thing is linear
Ahhh yes the classic right angle trick! (actually never seen this but it's awesome). Another reason that it could be lossy in the low end is that real-time frequency shifters work using allpass filters to (imperfectly) achieve a Hilbert transform, i.e. phase shifting each component by 90 degrees. But DC (0 Hz) cannot be phase shifted, and frequencies near 0 are also more poorly phase shifted by this method.
@@idiotinium I think maybe if you use a frequency shifter that's not intended for real-time use (one that performs phase shifts in the frequency domain via STFT, which introduces latency), it could work losslessly.
I think some of the phase shift at the end is possibly a result of imperfect frequency shifting; even if extremely accurate digitally, an ideal frequency shifter isn’t really possible afaik (at least in realtime)
This is similar to a thing i've had in mind that is it possible to flip the audio to a mirror image of the dynamics? When there's a peak that would be valley ect. Silence would be white noise or something. Just to have the dynamics upside down
Huh... Now this is interesting, it seems it's possible to reverse an audio sample _entirely in the audio domain_ which is probably useless but seems cool. Two flips gives you the reversed and frequency inverted version of the sample, so you just need to flip it on the frequency axis. I don't really know anything about convolution so I don't know if there's a simple way to use that, but you could amplitude modulate the sample with a fixed carrier frequency which will generate two mirrored sidebands as in an AM radio signal. Then you just low-pass out everything but the lower sideband and frequency shift back down by the magnitude of the carrier frequency, giving you the frequency-flipped version of the 180-degree-rotated convolved sample, which is the original sample played backwards. I have no idea what I'm talking about though.
Congratulations you just created the sounds inside the event horizon of a Schwarzschild black hole (i.e. time- and space-coordinates being swapped). Though I wonder how accurate this is, because you did it in the frequency domain, but who knows if the singularity of the black hole isn't just a Fourier operator at its heart. 😄⚛🌌🌠🌑
Interesting. This demonstrates that the time and frequency domains are naturally equivalent, as shown by the Fourier transform that works the same way both in forward and in reverse. I knew that turning a spectrum upside down is very simple, you just invert every other sample (i.e. modulate the signal by the Nyquist frequency), and this is fully reversible. One can also "skew" a spectrum by doing a convolution with a sweep signal. Doing a 90-degree rotation and getting back to the original signal after four rotations is another story. I wonder if there's a lossless way to do so.
theres definitely some way to do it completely losslessly, but idk if theres any tools that do that yet. not when they all have weird safeguards in like safety cutting nyquist or even frequency shifters straight up not having ranges that go to all the way nyquist freq
@@idiotinium I think I can write tools for that using my workflow. One would naturally do a Fourier transform of the entire sound clip, that effectively exchanges time and frequencies. The FFT outputs real/imaginary components, however, that should be somehow turned back into sound without loss of information.
I will die on this hill:
FL users are either the newbiest producers, clipping their masters at +5dB, or they are borderline NASA scientists. There is no in between😂
lmao this is true
wonder where i am
With my 10db master but okay music
Fax
Truest 😅
love it
"rotating audio 90 degrees" is the most insane thing ive seen a producer do
What about using photoshop to vocode something ?
@@trgrgtgrtgr7724 photoshop vocoder sounds so silly lmao
@@idiotinium”bro def used photoshop on this beat I can hear it”
@@idiotinium lmao It's actually a thing tho
photoshop@@trgrgtgrtgr7724 ? i just use "Photosounder Demo"
most insane part of this is seeing it done in FL studio and not matlab or something
Nice matlab shoutout. Never see these 2 worlds combine... well, not out in the wild
@@disivni any csv file with one column can be a wav file. any wav file can be a csv with one column. there IS common ground, someone just needs to put it to use.
Other producers been flipping samples for decades now but you are not like those other producers... you got your samples doing fucking cartwheels!
literally flipping samples rn
this is the craziest bit of sounddesign i have seen in ages
This is a geometric explanation of why taking the Fourier transform twice gives you the time domain back, only reversed. It's great to see another way of looking at it.
I do not produce music (yet) and this comment sounds like the most made up insider BS talk imaginable 😂
Like what are you gonna do to the time domain, space wizard?
@@vinceb8123 This is only really relevant if you like messing with frequencies. Fourier's transform is the best way to map time-domain bins into sine-space for further analysis
@@vinceb8123 if you just don't know how stuff works then just shut up and sit quiet
@@unrealenginerr5372 Yeah, your comment had way more value to the discourse! 🤡
you have no goddamn idea how ecstatic i am to learn that something i randomly conceptualized a long time ago is now actually possible
ive found myself making more IDM, glitch, noise, and botanica music and this is such an insane gamechanger for me
yoooo sick
sincerely congrats for figuring this stuff out, simply brilliant, maybe it seems to you that its goofy and gimmicky, or not very useful right now, but youre tapping in unexplored territory here and this could be the beginning of something bigger! ❤
This technique would work great for melodic riddim/color bass
agreed
Even engineering classes won't teach you this, hats off
formal education is silly
the classical formality really is silly; i’m actually hoping to develop a syllabus to teach introductory applied signal processing with music/audio. A lot of people seem to benefit much more from seeing something applied and trying it out before learning all the math behind it.
@@skrunglywungkus hell yeah thats awesome
This takes "sample flipping" to the next level
hell yeah im out here rotating samples
wow, the 270 degree one was cool. loved how it faded in all of those sounds
you can create some crazy distortion if you rotate it once, add some kind of delay, and then rotate it back. Could probably do some crazy sound design with this
Harmonics, yeah, that’s a great idea to explore
That's actually genius! What other effects could be useful swapping out the frequency domain for the time domain? Reverb to make stuff into weird white noise, I guess? Would EQ be like a particular volume automation curve over time? I'm not even sure what a flanger or phaser would I'm not even sure what a flanger or phaser would sound like, messing with all the frequencies in a sweep? Really cool idea to use the rotation like that!
@@NZsaltz perhaps this can be used to create distortion/saturation without aliasing? You'd just need a way to apply conventional 'distortion' in the time domain somehow... Wouldn't really be useful in a realtime setting, but probably useful in some way.
delay would be like layering a sound on itself with a frequency shifted version of itself multiple times in a stack
@@NZsaltz a flanger is actually caused using delay. Rotating it would sound pretty similar. A phaser, on the other hand, would depend on how many bands it had, but would be a weird comb-filtered delay I believe.
i want to try to turn this into a program so that it renders without latency and is lossless
that would be absolutely awesome
Yes please!
put an update here if you do this, because this is so sick but I don't have the software to do even this method myself rn
@@ekut1922 i started doing it. right now im looking into ways of making spectrogram. started with fft at first, but switched to Constant Q bcs its logarithmic (maybe ill make it possible to switch between them in the final version)
i have a MIGHTY NEED
Please make more videos like this
lmao sure if i discover anything equally or greaterly funky
This is so cool, I was so surprised that so little was lost after four rotations?! You could use this to hide a sneaky easter egg or ARG clue in a song :3
haha yeah it would be perfect for that
Absolutely insane, so cool to see you've figured it out
using sharex like that is actually the most genius part of this video
I was so bummed when you said it was purely a visual thing i cant believe you got this working
well it is a visual thing, youre seeing it do the thing in the spectrum lol
if its up to everyone to figure out uses for it
yesterday i thought of "oh you can hide a vocal watermark in a beat with this without it sounding too ass if you layer it with other stuff and speak very clearly"
@@idiotinium ohhhhh
I actually love the sound that the convolved slided sine wave does. Kinda like an all-pass filter
this is the most galaxy brain shit I've ever seen
I've been wondering about the possibility of flipping the frequency response of an audio source upside down for ages now but never had the know-how to do anything like this, thank you so much for satisfying that particular curiosity
Best video I've seen all week, my dude. Instant sub
dude i cant get over how unique this is, it's such an otherworldly sound design idea, that's awesome!! imagine if someone made a hybrid of this spectral rotation thing with granular synthesis... like a plugin which would take each grain and rotate them to some angle. maybe with some damping on high freqs to make them more controlled, but i honestly think this idea could spark a whole new plethora of ways to do sound design...
Thanks for showing us how to do this. I found out if you bitcrush a spectrogram rotated 90 degrees and rotate it back it creates a repeated granulized sound?? Im super confused why but it sounds cool.
hmm i guess it would do that, it'd be like if you repeated the original audio and layered that with a repeating version of it in reverse too
If you think about it, what any distortion does is just generate harmonics, in other words they generate information that is offset in the Y-Axis of a spectrogram. But if you rotate it, that information is now offset in the X axis - meaning you've generated information that is offset in time.
If I'm understanding it correctly, this means a delay plugin alongside this method could be used to generate harmonics in a novel way.
Harmonics become delay.
But you would need harmonic interval spaced delay for it to work. Or, hey, a novel road for non-harmonic tones
most insane thing i saw this week, congrats
this is so timely, I was just thinking how I wanted to do some spectrum range-shifting work (shifting, rotating, squashing, etc) the other day and have never done it before, now I have a few points to jump off at, cheers.
subscribed so fast theres not enough people getting this nerdy with it
Does copyright still apply if I rotate all my samples?
the courts cannot prove that you rotated a copyrighted song so no
@@Zetvue i would think the main problem would be anyone noticing, but if its noticed you could tell with enough knowledge that its just .. rotated
rotating something 90 degrees isn't a super lossy process so you aren't altering the original work THAT much, its a reversible process so its basically a novel way to store the signal weirdly
@@Zetvue yes they can
@@idiotinium if you rotated it 90deg, 180deg, or 270deg, no they can't
i was assuming they were just rotating it 90deg or something and not a full 360deg
This is what sound engineers should do 24/7, this is crazyy
at 5:25 where you do the phase shift test, the spectogram generates an eye, very cool
Seen that as well 😁
This is crazy please do more
this is amazing! i wont ever do it but it was great to watch how it's done and know it's posible!
Interesting! Im tired of videos how to compress drums… now this is new! I like it
Absolutely cursed, I love it
never seen this done in realtime before, every other thing like this exports the spectrum as an image, opening it in an editor, flipping it, then importing it back in. absolutley wild
I never comment, but this some of the coolest shit i have seen in a long time. Somehow YT algorithm works again
super slick stuff!
Oh, I thought this was impossible. I actually remember thinking about it for some practical use, cant remember which one :D Great stuff!
So sick 💙
Get better soon! 💛
This is like a brand new sentence but with lore behind it. I imagine cutting away the sweeping spund at the beginning and end contributes to the quality loss
yea i agree i also think so
Very much like the refractions of light seen in the glass of a window, a phone.
Even in a fucking fl studio, you can see the fundamentals of the universe. Man, that's fucking awesome.
lmao i never thought about it that way, thats very interesting
great work man, makes me think producers and programmers arent too different after all 💪💪
including mathematicians and whatever else, signal processing nerds, physicists etc!!!!
god this is so sick, my math and music brains are both going nuts over this. I really wanna try this now, but idk if i can in reaper
if you have a convolver and a linear frequency shifter its possiblə
There's a spectralist composer's career waiting to be founded on this technique
this is insane, thanks for sharing!
bro takes music out of it's dimensions and modifies it in 1 dimension above :D fkn great!
this is actually insane
I gotta learn this stuff this is crazy
The most insane thing I've learned on my own while playing with GNURadio Companion flowgraphs.
FFT itself does EXACTLY this. In one shot.
If you pass the studied signal through the rectangular-windowed FFT (I mean, a really large one, by the width of the whole audio length ceiled to 2^n instead of the default 1024 samples) and then interpret the result as the ordinary signal, you will get the spectrum rotated 90 degrees. Both the source and the result must be the I/Q signal with two channels (as we are operating with complex float32 values).
...Basically, Fourier transform just rotates the time-frequency domain 90 degrees counter-clockwise. Replaces peaks with sines, and sines with peaks. The idea is THAT simple.
...So, if you apply FFT 4 times in a row, the result will be the original signal as we just rotated the frequency domain 360 degrees.
lmao yeah
Awesome video thanks for sharing
i love videos like this! :)
To anyone that wants an easier way of doing this, I recommend photosounder, with which you can edit audio as if it were a picture, which includes layering, rotating, changing contrast etc.
do not do this if you're trying to have a good outcome lol, you will have no phase spectrum and quantize your amplitude bins to the bit depth of the image you're saving. additionally think about the pixel interpolation by reading the image and the window function used to generate the spectrum image, also window overlapping etc. that process is supremely lossy
Trippy. No idea that could work consolver n stuff. Interesting stuff
That is brilliant!
I imagine you could do something similar if you manipulate the spectrum with an image editor and a tools Photosounder or Coagula, but it's amazing to do it strictly in the audio domain, so to speak! A lot of potential for interesting sound design, and of course Easter Eggs :)
keep in mind the phase spectrum is a very critical part of any signal's fourier decomposition, most of those tools only export magnitude/amplitude spectra, also with no window overlap or anything (some even distort it such that it no longer has a linear frequency axis). plus the partials' amplitude will be quantized to the bit-depth of the image further losing information
keeping it in audio form retains phases and stuff cus its the pure actual math on the signal
It sounds really cool honestly, some stuff aphex twin would make
That's so cool. I wish you'd played what the original audio sounded like though
I can think of exactly 0 use cases for this other than maybe encrypting something like someone said in a comment below, but it sounds sick.
How did you even find out how to do this? I have no understanding of why what you did achieved the result.
Edit: nvm, I watched your previous video about this, and it makes a lot of sense, this is insane lol.
"it sounds sick" is a use case if you think about it
i saw a recent matt parker (standupmaths) video explaining how a rotation is equivalent to 3 skews if you use specific ratios for the skews, and i also noticed that convolving with a sine sweep was one way to extremely offset phases, and a frequency shifter skews vertically when automated linearly so yeah
also it would be a cool way to hide a secret in some puzzle thing for sure!
you can use frequency shifter from spectral pack. it has higher range than khs iirc
do you have a link to that?
awesome swag sauce
:D
This would be interesting to do at a sample rate of say 20khz or 24khz so when it's inverted and backwards after 180deg rotation the majority of the audio is still within the audiable range.
This is super interesting.
I have a few ideas with this:
Harmor can actually process images to generate sound.
Also, Melda’s MCabinet is a Convolver that only takes tonal information and not phase/time
So it could provide a different result
Could the 180 rotation be used in combination with other effects to obtain a "negative harmony" effect?
maybe, i wouldn't count on it since it's linear instead of how notes are a geometric progression instead of arithmetic
notes have an exponential scale while this thing is linear
HOLY SHITE THISIS SO CLEVER
The first rotation might sound cool as a convolution. This is really cool though! Sub'd!
these are extreeeemly neat gimmicks i will 100% be using, but what...is this track it sounds sick 👀
upcoming song, will release it under my actual artist accounts of course. still trying to think of a name for it that suits it tho
this is insane
oh i gotta try this sometime as another step in making cool pad samples, with some processing in the middle
ok so followup, i couldn’t get it to rotate exactly 90 degrees, always a little off
and the latency was testing my patience real hard lol
but it was fun to play around with
THIS IS SO COOL
this is so cool
Whhaaaaat the friiick I crapped my pants when you played the last one
If i have seen some hame community nowadays, I'd say that this will DEFINITELY be used for an ARG sometimes in the future.
that lit bro.
So cool!!
This is so insane, only complaint is that your tutorial style is absolutely adhd repellent
lmao, i have adhd
Untrue my friend, I have raging adhd and I had 0 problems following our boy hete do his thing.
:3
Bro, that's hella cool. You need to make a patcher plugin that allows you to rotate audio
This could make sense as a technique for the soundtrack of a work about time travel.
man sometimes I think I've seen producers do it all, but then I come across a producer doing some shit like this
aight that's pretty cool
thank you
phase shift might come from an anti-aliasing LPF (near nyquist) used in the rendering process
yeah and the low end dropping out too after 4 rotations i suspect
@@idiotinium right
Ahhh yes the classic right angle trick! (actually never seen this but it's awesome).
Another reason that it could be lossy in the low end is that real-time frequency shifters work using allpass filters to (imperfectly) achieve a Hilbert transform, i.e. phase shifting each component by 90 degrees. But DC (0 Hz) cannot be phase shifted, and frequencies near 0 are also more poorly phase shifted by this method.
theres gotta sooooome way out there that works tho, surely
@@idiotinium I think maybe if you use a frequency shifter that's not intended for real-time use (one that performs phase shifts in the frequency domain via STFT, which introduces latency), it could work losslessly.
I think some of the phase shift at the end is possibly a result of imperfect frequency shifting; even if extremely accurate digitally, an ideal frequency shifter isn’t really possible afaik (at least in realtime)
im speechless wft wow
This is similar to a thing i've had in mind that is it possible to flip the audio to a mirror image of the dynamics? When there's a peak that would be valley ect. Silence would be white noise or something. Just to have the dynamics upside down
You're a fucking maniac I love it so much thank you
youre insane man, that's fucking awesome
wow crazy
I like I like I like !!
pure black magic
rotating audio 270 degrees can produce some cool effects, kinda like the THX logo in a way
you could actually do audio encryption with this. The key culd be the degrees rotated
Huh... Now this is interesting, it seems it's possible to reverse an audio sample _entirely in the audio domain_ which is probably useless but seems cool.
Two flips gives you the reversed and frequency inverted version of the sample, so you just need to flip it on the frequency axis. I don't really know anything about convolution so I don't know if there's a simple way to use that, but you could amplitude modulate the sample with a fixed carrier frequency which will generate two mirrored sidebands as in an AM radio signal.
Then you just low-pass out everything but the lower sideband and frequency shift back down by the magnitude of the carrier frequency, giving you the frequency-flipped version of the 180-degree-rotated convolved sample, which is the original sample played backwards.
I have no idea what I'm talking about though.
lmao yeah i was looking into that recently actually
amazing :D
What does the 180 degrees flipped version sounds like if transposed down to a lower/normal frequency domain?
I was like, "who cares?" and then I watched it. Pretty cool waveform wizardry
this is crazy
did 6 frequency shifters instead of 5 so its equal on the macro. feel like that might clean it up a little
i think i just gotta find a clean enough linear frequency shifter vst that would let me go up to nyquist lool
Congratulations you just created the sounds inside the event horizon of a Schwarzschild black hole (i.e. time- and space-coordinates being swapped). Though I wonder how accurate this is, because you did it in the frequency domain, but who knows if the singularity of the black hole isn't just a Fourier operator at its heart. 😄⚛🌌🌠🌑
frequency is inverse time, not a spatial dimension
soooo im swapping time and inverse time
Interesting. This demonstrates that the time and frequency domains are naturally equivalent, as shown by the Fourier transform that works the same way both in forward and in reverse.
I knew that turning a spectrum upside down is very simple, you just invert every other sample (i.e. modulate the signal by the Nyquist frequency), and this is fully reversible. One can also "skew" a spectrum by doing a convolution with a sweep signal. Doing a 90-degree rotation and getting back to the original signal after four rotations is another story. I wonder if there's a lossless way to do so.
theres definitely some way to do it completely losslessly, but idk if theres any tools that do that yet. not when they all have weird safeguards in like safety cutting nyquist or even frequency shifters straight up not having ranges that go to all the way nyquist freq
@@idiotinium I think I can write tools for that using my workflow. One would naturally do a Fourier transform of the entire sound clip, that effectively exchanges time and frequencies. The FFT outputs real/imaginary components, however, that should be somehow turned back into sound without loss of information.
i was trying to figure out this a while ago and then got too lazy, guess someone did it first haha :D