Split Thread henryco's new paper

[The Big Dog]

Let's be clear.

I don't intend to try to convince the JREF's average Joe of anything. I would be a fool to try to. No one ever convinced you guyz of anything. I am not a magician.

Now some people here provides arguments and most of the time oversell them. What I can do with that is explain why the arguement will not convince a neutral, non zealot crowd (ie somenone neither from JREF nor Infowars). Which I did as would anyone with *real* experience of audio editing.

That being said, I don't need to get into details. I don't mind if you guys still think it's a good argument. You want to post it in a more open forum and look like an idiot ? Please be my guest.

My point is just that I don't want to let any doubt that 240... audio demo is crap. And that all of you who just can't zoom the image in your browser and compare the two enveloppe for differences are juste a joke. It tool me 30 sec. How hard can this be ?

That is some fancy alternative English you are using there, sport.

Now, as an expert in "audio engineer" I need you help. I blew it up 4000% and all I can see in an individual pixel. Thank you for coming back after over a year with your earth shattering dicovery to make "fun of people."

Sorry you don't actually have the time to do anything more.

 

[grandmastershek]

So the courts don't work well for us. (Yet).

And that will be when courts abandon all standards of logic and evidence.

 

[George 152]

That is some fancy alternative English you are using there, sport.
.

Hey, you try to type in a straight jacket

 

[Redford]

Now, as an expert in "audio engineer" I need you help. I blew it up 4000% and all I can see in an individual pixel.

As I said, zoom enough to make the comparison easier. 400% is doing it for me, but it depends on your screen, browser, etc. Basically the game is like "find the 7 differences between the two pictures". If that was copy and paste, there shall be none and you should be able to find a way to superpose the two envelopes exactly, pixel per pixel. Or at least (as this depend on how the user interface is displaying envelope), event for event : here a high peak followed at a distance by two peaks exactly of the same amplitude on both envelopes, that sort of thing.

By doing this you'll quickly realize that the two envelopes may look alike at a distance (which is expected if real) but are two separate noises.

Note that it doesn't prove it hasn't been edited. But not by a copy and paste.

 

[carlitos]

Let's be clear.

I don't intend to try to convince the JREF's average Joe of anything. I would be a fool to try to. No one ever convinced you guyz of anything. I am not a magician.

Now some people here provides arguments and most of the time oversell them. What I can do with that is explain why the arguement will not convince a neutral, non zealot crowd (ie somenone neither from JREF nor Infowars). Which I did as would anyone with *real* experience of audio editing.

That being said, I don't need to get into details. I don't mind if you guys still think it's a good argument. You want to post it in a more open forum and look like an idiot ? Please be my guest.

My point is just that I don't want to let any doubt that 240... audio demo is crap. And that all of you who just can't zoom the image in your browser and compare the two enveloppe for differences are juste a joke. It tool me 30 sec. How hard can this be ?

This is your claim. Perhaps you could post evidence for your claim?

Print the screen. Show your work.

As I said, zoom enough to make the comparison easier. 400% is doing it for me, but it depends on your screen, browser, etc. Basically the game is like "find the 7 differences between the two pictures". If that was copy and paste, there shall be none and you should be able to find a way to superpose the two envelopes exactly, pixel per pixel. Or at least (as this depend on how the user interface is displaying envelope), event for event : here a high peak followed at a distance by two peaks exactly of the same amplitude on both envelopes, that sort of thing.

By doing this you'll quickly realize that the two envelopes may look alike at a distance (which is expected if real) but are two separate noises.

Note that it doesn't prove it hasn't been edited. But not by a copy and paste.

 

[W.D.Clinger]

This is your claim. Perhaps you could post evidence for your claim?

Print the screen. Show your work.

For reasons explained below, printing the screen and showing his work is unlikely to help.

Now some people here provides arguments and most of the time oversell them. What I can do with that is explain why the arguement will not convince a neutral, non zealot crowd (ie somenone neither from JREF nor Infowars). Which I did as would anyone with *real* experience of audio editing.

My experience with audio editing, computer graphics, and signal processing is modest but real enough.

My point is just that I don't want to let any doubt that 240... audio demo is crap. And that all of you who just can't zoom the image in your browser and compare the two enveloppe for differences are juste a joke. It tool me 30 sec. How hard can this be ?

It looks impossible to me. The image posted by 240-185 appears to have been captured off the screen from a display produced by some audio editing program. If so, it already contains at least one and possibly two or three generations of pixelation/quantization error: one when the audio waveform is converted into two graphical forms by the audio program, and possibly a second generation when the graphical output of the audio program is rescaled by the computer's window system, and possibly a third generation if 240-185 rescaled the image again for display at JREF. That's on top of any registration errors 240-185 may have introduced when he tried to line up the two waveforms.

No one who understands computer graphics would expect the two waveforms in the JREF image to look exactly the same, even if they are representations of exactly the same audio waveform.

By doing this you'll quickly realize that the two envelopes may look alike at a distance (which is expected if real) but are two separate noises.

Note that it doesn't prove it hasn't been edited. But not by a copy and paste.

It looks to me as though the differences between those two waveforms are small enough to be explained as above. If the two waveforms are actually independent, then it's a truly remarkable coincidence that they look so similar.

 

[carlitos]

Ah cool. I couldn't have been bothered to sample explosion audio and analyze the wave form. I'd much rather work on real music. At least it serves a purpose.

 

[W.D.Clinger]

Ah cool. I couldn't have been bothered to sample explosion audio and analyze the wave form. I'd much rather work on real music. At least it serves a purpose.

By coincidence, I taught a course last semester in which the semester project was to write a computer program that, given a short excerpt of music in the form of a .wav or .mp3 file, searches a longer .wav or .mp3 recording to determine whether the excerpt matches any portion well enough to conclude that the two recordings were both derived from some common ancestral recording.

That's an interesting problem because the two input recordings may have been derived via completely different histories of encoding, compression, and other manipulations that disguise their common origin. Even so, it is possible to determine whether they have a common origin with high reliability.

And that just happens to be the question here, except we're working with the extremely low-resolution graphical representations posted by 240-185 instead of with the actual recording he was using.

ETA: With the above as background, I might as well go into more detail about this:

Note that it doesn't prove it hasn't been edited. But not by a copy and paste.

That remark suggests that Redford is thinking about the audio editing that goes on when you have a nice clean digital studio master, or at least a WAVE file. IIRC this audio came from a video, so it had already been subjected to fairly severe MPEG (or similar) lossy compression.

Even if the two segments in question had been identical in some uncompressed version of the audio, lossy compression algorithms will compress them a little differently depending on where they happen to fall with respect to the windows on which digital filters and discrete Fourier transforms of the lossy compression algorithm are applied.

That means we should expect some small differences between the segments, even if they were identical before compression, before the graphics software added further differences.

That's why audio matching almost always uses algorithms for approximate matching instead of the more straightforward exact matching.

 

[beachnut]

I am coming after the battle but I just wanted to say that the sound demonstrated to be "mono" and "copy pasted" by 240-... are actually stereo and two different sounds. Just zoom the enveloppe in your browser and compare peaks from both sound : not the same (or you can use an image editor with layer transparancy to see that they don't fit).

I find it amusing to see everyone congratulate him and take for proven what's an especially cheesy argument. Pretty typical JREF : you guys come here to hear what you want/need to read. I see other "audio" post here, a matter I am a bit aware of. If you think there are other points that need double checking for you let me know - I am not patient to read everything and see if there's something else than nonsense.

What is your point? Thermite don't make explosive noises? No thermite was used? No explosives were used on 911? What is the point?

North Tower: A Too Fast Collapse for Gravitational Collapse Theories

... the topic, and it is a moronic delusion based on nothing but woo. So what was your point, and how does it relate to the failed topic? Any math to go with this stuff?

 

[twinstead]

I have to agree with W.D. Clinger--it would be truly remarkable that two independent waveforms would be so similar.

 

[Redford]

It looks impossible to me. The image posted by 240-185 appears to have been captured off the screen from a display produced by some audio editing program. If so, it already contains at least one and possibly two or three generations of pixelation/quantization error: one when the audio waveform is converted into two graphical forms by the audio program, and possibly a second generation when the graphical output of the audio program is rescaled by the computer's window system, and possibly a third generation if 240-185 rescaled the image again for display at JREF. That's on top of any registration errors 240-185 may have introduced when he tried to line up the two waveforms.

No one who understands computer graphics would expect the two waveforms in the JREF image to look exactly the same, even if they are representations of exactly the same audio waveform.

Actually one could, because for both optimization and design reason that's the best way to do it. You want to avoid calculating twice the representation of a sound ; if you cut one in a separate bloc you will not redraw its display (neither each time you move it, shorten it, etc.), you'd rather get your representation by copying it from the original clip too. So in any audio editor if you copy a clip its envelope will remain exactly similar in the new clip, even if you loop it.

That being said, zooming in and out may desynchronize the two rendering (depending on how it's done) - that's why I suggested comparing peaks and amplitudes and not just pixels. Even if pixel per pixel rendering change slightly, data at this level will remain valid (after all, that's the info you expect from an envelope).

BTW the software he uses here is Adobe Audition, which I assume everyone with some audio editing experience should be able to tell.

That remark suggests that Redford is thinking about the audio editing that goes on when you have a nice clean digital studio master, or at least a WAVE file. IIRC this audio came from a video, so it had already been subjected to fairly severe MPEG (or similar) lossy compression.

More or less. I can imagine several reason why, if done well, you couldn't probably tell any edit just by staring at an envelope. Some are just basic editing practices (doing this well isn't exactly rocket science, you just have to take in & out points at zero level), some are mixing consideration, some, as you state properly, are compression artefacts that may "soften" the edit.

It looks to me as though the differences between those two waveforms are small enough to be explained as above. If the two waveforms are actually independent, then it's a truly remarkable coincidence that they look so similar.

Take a microphone and record a firework exploding. Do it again, with the exact same fire work in a similiar location, several time. Compare : all envelopes will look alike in the same way those above are. I don't know why you expect anything else : similar noises have very similar envelopes. Envelopes depends on the noise, it's not fingerprints.

This is your claim. Perhaps you could post evidence for your claim?

I think I did. Did you do the comparison I suggested ?

 

[W.D.Clinger]

Actually one could, because for both optimization and design reason that's the best way to do it. You want to avoid calculating twice the representation of a sound ; if you cut one in a separate bloc you will not redraw its display (neither each time you move it, shorten it, etc.), you'd rather get your representation by copying it from the original clip too. So in any audio editor if you copy a clip its envelope will remain exactly similar in the new clip, even if you loop it.

You're forgetting that the lower envelope came from a different part of the audio recording, so the audio editor doesn't know that the upper envelope is the same as the lower.

You're also forgetting that the two envelopes are different, because MPEG's lossy compression would have introduced differences between the two excerpts even if there were no differences between the excerpts in the uncompressed version.

Take a microphone and record a firework exploding. Do it again, with the exact same fire work in a similiar location, several time. Compare : all envelopes will look alike in the same way those above are. I don't know why you expect anything else : similar noises have very similar envelopes. Envelopes depends on the noise, it's not fingerprints.

You would learn something from doing that exercise yourself.

I have already done that exercise in a different context: comparing different recordings of the same piece of music, performed by the same performer, on the same instrument, with the same recording setup. The articulations, which are fairly noisy, often produce visible differences between the envelopes, as do the phases of low-frequency components.

Welcome to JREF, by the way.

 

[Redford]

You're forgetting that the lower envelope came from a different part of the audio recording, so the audio editor doesn't know that the upper envelope is the same as the lower.

Indeed! You're right.

You're also forgetting that the two envelopes are different, because MPEG's lossy compression would have introduced differences between the two excerpts even if there were no differences between the excerpts in the uncompressed version.

Not on this. Compression would affect thing at a way smaller scale than what we see here.

You would learn something from doing that exercise yourself.

I think I already did. But I also payed attention at the scale ;-)

The articulations, which are fairly noisy, often produce visible differences between the envelopes, as do the phases of low-frequency components.

It's entirely different from what I suggested, ie making the same noise. A musical instrument is not as dull as a constant amount of explosive in a constant place - well, not all of them at least. That being said, hit a triangle as hard as you can several time - my bet is that you wouldn't see or hear differences.

Welcome to JREF, by the way.

Not entirely my first time here, although I am not a regular. Thanks for welcoming, thus.

 

[W.D.Clinger]

You're forgetting that the lower envelope came from a different part of the audio recording, so the audio editor doesn't know that the upper envelope is the same as the lower.

Indeed! You're right.

You're also forgetting that the two envelopes are different, because MPEG's lossy compression would have introduced differences between the two excerpts even if there were no differences between the excerpts in the uncompressed version.

Not on this. Compression would affect thing at a way smaller scale than what we see here.

My point was redundant and I failed to explain it clearly: Even if the audio editor were so bizarre as to compare the two excerpts with an eye toward rendering them only once if they were identical, they would still be rendered separately because they aren't identical at the fine scale.

You would learn something from doing that exercise yourself.

I think I already did. But I also payed attention at the scale ;-)

You think you already did?

If you have actually done the exercise you suggested, then you can show us the results.

 

[Redford]

You think you already did?

If you have actually done the exercise you suggested, then you can show us the results.

Actually 240 ... said "I did prove my point because those two envelopes are identical".

I said "nope, it doesn't, compare accurately peaks by rythme and amplitude".

Some of you seem to think that for some reason I shall prove them what can be checked by anyone doing what I did in 10 seconds.

Someone who would want to know would just spend this 10 secs. So I assume that as expected here you guyz aren't the "I want to know" type. Instead you're spending way more time into making arguments to explain why you're right not to want to know. I understand that defines your type, then.

So this mean that if I spend 15 minutes making an even more telling demo I would not get a reaction of someone telling me if I answered his/her question. I would get reaction of people explaining me why they wouldn't even consider the demonstration.

I am sure all of you guys here can reckon a pattern that is kind of the wallpaper here (although none would admit, because, you know, you're from that type). Well, not for me. As stated, I just stop my point where I have disproven a point. Now if someone wants to disprove me, he she (ok he) will have to work a little. And probably realise he can't by doing so.

 

[W.D.Clinger]

Actually 240 ... said "I did prove my point because those two envelopes are identical".

I said "nope, it doesn't, compare accurately peaks by rythme and amplitude".

Some of you seem to think that for some reason I shall prove them what can be checked by anyone doing what I did in 10 seconds.

It took me a little more than 10 seconds, but I confirmed your statement that the envelopes were slightly different. I went beyond that confirmation by noting that the envelopes are remarkably similar.

Someone who would want to know would just spend this 10 secs. So I assume that as expected here you guyz aren't the "I want to know" type. Instead you're spending way more time into making arguments to explain why you're right not to want to know. I understand that defines your type, then.

You, on the other hand, base your assumptions upon your expectations.

So this mean that if I spend 15 minutes making an even more telling demo I would not get a reaction of someone telling me if I answered his/her question. I would get reaction of people explaining me why they wouldn't even consider the demonstration.

So you recommended an exercise that would take far longer than 15 minutes (if only to purchase the firecrackers!), but you are unwilling to spend that time yourself.

You also dismiss my experience, which includes designing, writing, and testing computer programs that can distinguish recordings derived from a common ancestor from other recordings of the same music made by the same performer on the same instrument on the same day using the same recording equipment, studio, and setup.

I am sure all of you guys here can reckon a pattern that is kind of the wallpaper here (although none would admit, because, you know, you're from that type). Well, not for me. As stated, I just stop my point where I have disproven a point. Now if someone wants to disprove me, he she (ok he) will have to work a little. And probably realise he can't by doing so.

Sounds as though you're determined to conclude that your bias has been confirmed.

 

[Redford]

It took me a little more than 10 seconds, but I confirmed your statement that the envelopes were slightly different. I went beyond that confirmation by noting that the envelopes are remarkably similar.

Well, 240... point was not that they were just slightly different. So my point doesn't really need more than showing he's wrong in that. That being said, I also said the "remarkably similar" is what I expect for two explosions due to the same thing in a similar acoustic context. Now if you think envelopes of two similar sounds wouldn't be similar, I suggest you try to prove your point by recording two sounds that are identical while strong differences in their envelopes, because that's an interesting theory.

You, on the other hand, base your assumptions upon your expectations.

You may have noticed I am not a native english speaker so beating me with words is like kicking a puppy, you know...

So you recommended an exercise that would take far longer than 15 minutes (if only to purchase the firecrackers!), but you are unwilling to spend that time yourself.

What do I gain ?

You also dismiss my experience, which includes designing, writing, and testing computer programs that can distinguish recordings derived from a common ancestor from other recordings of the same music made by the same performer on the same instrument on the same day using the same recording equipment, studio, and setup.

I don't think I did that, and if I did I apologise. The "Adobe Audition"'s bit maybe ?

Sounds as though you're determined to conclude that your bias has been confirmed.

Well it's a rethorical trap and I won't blame you because I started it. Let's just consider that I explained why I don't want to step in what I feel is a dishonnest "proove your point" strategy. In your case, it's a bit different thus as you admit the differences, which is kind of my point, really. The argument has changed from "audio edit proven" to "audio edit not proven but not proven wrong". Although not entirely for the same reasons it also is my position and I think FHC's one. So basically why split hair on the topic ?

 

[DGM]

So in any audio editor if you copy a clip its envelope will remain exactly similar in the new clip, even if you loop it.

I have to ask. What does "exactly similar" mean?

 

[W.D.Clinger]

You also dismiss my experience, which includes designing, writing, and testing computer programs that can distinguish recordings derived from a common ancestor from other recordings of the same music made by the same performer on the same instrument on the same day using the same recording equipment, studio, and setup.

I don't think I did that, and if I did I apologise. The "Adobe Audition"'s bit maybe ?

No, I was referring to your dismissal of my experience on the grounds that musical instruments are not explosives.

I also enjoyed the way you parried my remark about the "phases of low-frequency components" by suggesting I "hit a triangle as hard as you can several time". I have a counter-suggestion: more cowbell.

In your case, it's a bit different thus as you admit the differences, which is kind of my point, really. The argument has changed from "audio edit proven" to "audio edit not proven but not proven wrong".

Agreed.

Anyone who really wants to pursue this argument further should go back to the audio recording instead of arguing about a low-resolution screen grab.

 

[Redford]

No, I was referring to your dismissal of my experience on the grounds that musical instruments are not explosives.

I also enjoyed the way you parried my remark about the "phases of low-frequency components" by suggesting I "hit a triangle as hard as you can several time". I have a counter-suggestion: more cowbell.

It's a misunderstanding then - I was literal and didn't mean to dismiss your experience. My own experience comes from the music software industry and I was thinking to that particular field of instruments emulation, whether you do it sample based or synthesis way (or both). Deciding which way to go includes studying samples of an instrument and tracking the variations and trying to guess what cause them. Triangle - or cowbell - are the simplest thing : assuming you hit the same part with the same stick at the same strenght, you'll have the same sound (and same enveloppe). A guitar is way more complex - it shall be noted thus that here you rather see the limit of amplitude envelopes. Two similar envelopes can still sound noticeably different. That's because amplitude envelope just show a part of what defines a sound. Ok, this is really OT...