I discovered a sound that leaks your exact location (and it's everywhere)

I have found off-grid structures in the middle of a rainforest, things 600 meters deep in the ocean and some stuff I can't even share on YouTube. But what

I'm going to share today is no exaggeration the single most powerful geoloccation clue I've come across. And

to illustrate this, I'm going to geoloccate this. It's a video, but

geoloccate this. It's a video, but nothing really changes visually. If you

want to pause and try it yourself, it's linked in the description. The only

information you have is [music] that it was filmed September 8th somewhere in the continental US. That's all you get.

But with that information, [music] it is possible. And I'm going to show exactly

possible. And I'm going to show exactly how.

So, I did a video geoloccating with thunder sounds. The benefit is that they

thunder sounds. The benefit is that they travel really far and they penetrate structures extremely well. The downside

is that we were only able to get about 5 km precision. [music] that could tell

km precision. [music] that could tell you basically what city it was filmed in and maybe what side of town. What we're

using today has those same benefits, but here's the precision.

So, I'm going to play you two sections of audio. One has the information

of audio. One has the information required to find this location. The

other does not. See if you can tell which one does.

You probably couldn't tell because it's completely inaudible, but if I open this in a spectrogram here, we can see all the frequencies. So, if you're not

the frequencies. So, if you're not familiar with a spectrogram, the xaxis is time and the y-axis is frequency.

Now, even with this, you still really can't see anything. It's only when I zoom into the right frequency and then increase the resolution that you start

to see it. I'll do one more.

This is a Doppler curve from an aircraft, and it's what we're going to use to locate this video.

Now, fair warning, this is insanely powerful, but it also requires a lot of complex stuff. This took me months to

complex stuff. This took me months to build, over 5,000 lines of code. I was

working with 50 GB Park files, 3D physics modeling, countrywide DEMs, and so much more. That's all to say. This is

going to be a long video, but I'm going to explain every step in a way that anyone can understand it, even if you know nothing about any of this stuff.

So, with that out of the way, let's look at the rest of this audio. We have one Doppler curve here. If I keep going to the right, you can see another one. It's

a lot louder, and you can also see the harmonics on top. I'll zoom out a bit. I

think we have one more here. So, these

two are what we'll be using. But, let's

take a step back. Why do we even care about Doppler curves? What are we going to be using these for? So, there's

something called ADSB data, and it's data on pretty much every plane in the sky. Its position, speed, altitude, and

sky. Its position, speed, altitude, and a bunch of other things. So, my first thought was, could I narrow this down to only places that heard a plane, and then

920 seconds later another one? To do

that, we need all this ADSB data first.

They have an API you can pay for.

There's also a Python package just to scrape the data, but it's massive. So, I

ran this for a few hours to get all the flights in the continental US on September 8th. Now, here's what that

September 8th. Now, here's what that data looks like. By default, it downloads the entire flight path. That's

why you see things going to Hawaii over here or other countries. But anything

that flew over the US on that day is in this data set, including the two planes in our recording. So, now we just need to find them. So, like I said, I want to narrow down to the locations that heard

a plane and then 920 seconds later another one. Now, these Doppler curves

another one. Now, these Doppler curves are made by a plane, of course, but specifically a propeller plane. I

boosted some frequencies so you can hear it better. Let me play you a bit of the

it better. Let me play you a bit of the louder one, and you can hear pretty intuitively that it's a propeller plane.

One of the fields in the ADSB data is type code. Now, this doesn't say

type code. Now, this doesn't say propeller versus jetplane, but I did some research, found this lookup table, and this does. Now there's some missing data in that lookup table. So instead of filtering to the propeller planes, I

just filtered out all the known jet types. But after doing that, the map

types. But after doing that, the map looks like this.

But now time for this time offset filter. So I'll do this on a small area

filter. So I'll do this on a small area to illustrate first. This pink line is the ADSB data for one flight. You can

see the individual flight points here where we have coordinates, altitude, time stamp, things like that. What we do is lay a 200 meter grid of points and then for each point note the time that

it would have heard this flight. Now

it's time for our first new terminology.

CPA, not that one. Closest point of arrival. So for an observer, say here,

arrival. So for an observer, say here, CPA is where the plane is closest to that observer. The reason that's

that observer. The reason that's important is when we say heard a plane 920 seconds later, hearing a plane is not really a single point in time. If

you're standing at this point, you hear the plane here and here and here. This

whole Doppler curve is almost 2 minutes.

So, what we need to use is the CPA time, which is roughly the middle of this.

It's when the plane was closest and loudest. So, for this flight, lay a 200

loudest. So, for this flight, lay a 200 m grid of points. I did a pretty conservative 10 km radius. Like, beyond

that, you probably wouldn't hear a plane even in like ideal conditions. So,

that's just make sure we don't miss anything. And then for each one of these

anything. And then for each one of these points, we note the CPA time for this flight. Two important things here. First

flight. Two important things here. First

is that CPA time actually means the time when the plane was closest. For this

example, that would be probably this point. So 22 46 36. That's actually not

point. So 22 46 36. That's actually not what we need, though. Now, for the second terminology, what we need is the delayed CPA time. It's the time that the CPA would have been heard by an

observer. And that's because the sound

observer. And that's because the sound has to travel from here all the way 3 plus kilometers to a listener. That was

the first important thing. The second is that planes are in the sky. Meaning the

distance that we use in this calc needs to be the slant distance, not just the horizontal distance. Luckily, the ADSB

horizontal distance. Luckily, the ADSB data gives the plane's altitude.

However, altitude is not the height above the ground. Altitude is the height above sea level. So to get this slant distance, we actually need to know the elevation for each grid point. And to do

that, we'll use a DEM. That'll tell us at any point the elevation.

So now with that we can actually say for our point here for this flight the CPA time of 224634

but the delayed CPA time of 224646.

So we do this for every point here and that's one flight done. Then we just do that for all flights in the area and we're left with this grid of points because basically every point heard at least one flight especially since I gave

the 10 km radius. But for each of these here, let me zoom in on one. For each of these, what I have is a list of flight IDs and the associated delayed CPA times. Basically, if you were standing

times. Basically, if you were standing here on this day, at each of these times, you could look up and you could hear a plane. So, back to our goal. What

we want is somewhere that heard a plane and then 920 seconds later heard another one. So for each of these points, we

one. So for each of these points, we then just check for that offset. Here's

what that actually looks like. So here's

an example point and each of those delayed CPA times. And then we have a 920 second offset window. And we just look here. Nothing. 920 seconds later.

look here. Nothing. 920 seconds later.

This one again. Not quite. And then this one over here. There we go. Now, when we say 920 seconds apart, it matters where we're measuring from. Like are you calling the center here or here or here?

And that's like a plus or minus 5second variance. So that's why if you look

variance. So that's why if you look closely on this, there's a bit of wiggle room that we're allowing for in that 920 second offset. But anyway, that's the

second offset. But anyway, that's the methodology. So how many of our points

methodology. So how many of our points survived that filter? It's about 35,000.

And that's for this small test area, which is less than 0.05% of the continental US land area that we need to search. Like just a side tangent, I don't think people understand

how much land area there is in the US.

If you divided the continental US into a bunch of football fields and laid them end to end, it would wrap all the way around the Earth and to the moon and

back almost 200 times. That's what we're searching for this. So, we need to do a lot better and we can, but we need to

know some things about Doves.

All jokes aside, this is actually so cool. Doppler curves will give us a ton

cool. Doppler curves will give us a ton of additional information and I actually don't need to teach you anything because you already know this intuitively. Let

me play you two clips.

You can already tell that clip 2 was going faster. The reason is because the

going faster. The reason is because the drop in frequency was much larger. And

that drop in frequency can actually tell us the exact speed that an object was going. So assuming linear constant

going. So assuming linear constant velocity, there are only two things that determine what this Doppler curve looks like. The velocity determines how big of

like. The velocity determines how big of a frequency drop. And our CPA distance, remember that one, determines how quickly that drop occurs. So for the Doppler curve in our audio, we can fit a

curve to that and know how fast the aircraft was going and how far away it was from the listener. So now instead of filtering to points that hurt an aircraft and then 920 seconds later

heard another one, we can filter to points that heard an aircraft going a certain speed at a certain CPA distance and then 920 seconds later heard another plane going a certain speed at a certain

CPA distance. That's a lot more

CPA distance. That's a lot more constraining. So let's measure one of

constraining. So let's measure one of the Doppler curves. You can see that the velocity makes the frequency drop larger and then the CPA slant distance just changes how quickly that that drop occurs. What you'll notice on this one

occurs. What you'll notice on this one is that it doesn't fit a perfect Doppler curve though like this right hand side does. Left hand side not so much.

does. Left hand side not so much.

That'll be important later. But what we do in the meantime is we just take a pretty generous range for velocity so that we don't accidentally filter anything out. So here's kind of a decent

anything out. So here's kind of a decent fit on the right side at 100 m/s. I

can't confidently say this plane was going exactly 100, but we can be pretty sure it wasn't going 70. Similarly for

the time offset, there's still a couple seconds of like potential measurement error here. Okay, best match 15115. I'll

error here. Okay, best match 15115. I'll

move it to 15113. Slide that fundamental frequency up a little bit. And like,

yeah, you know, that could work. That's

just to say we still can't pin down the offset time precisely to the second. But

we do this for event one and event two, and we get the measurement ranges for each. So, the first thing is the speed

each. So, the first thing is the speed filters. Here are all of the prop

filters. Here are all of the prop flights, but here are all the ones matching the event one velocity range.

And then in pink, here are the same ones before event two. So now what we're looking for is a point that heard a green line and then 920 seconds later heard a pink line. But remember, the

other thing we know about this too, in addition to the velocity, is the distance away. So on this green line for

distance away. So on this green line for event one, we don't need to lay that entire grid from 0 to 10 km away. We

know the pretty precise slant distance that the observer was from each of these. So, here are those points.

these. So, here are those points.

They're about 73,000.

But if I zoom into any of these clusters, what do you notice?

Airport.

Another airport. Most of those 73,000 points are just near airports. And it

makes sense cuz that's where the planes are. More air traffic just means higher

are. More air traffic just means higher chance that you happen to have two flights that meet those criteria. But

the signal in our video is not only what we hear, it's also what we don't hear.

There's no Doppler curve between these two. That's almost 15 minutes of not

two. That's almost 15 minutes of not hearing a plane. You're not going to get that right outside of a major airport.

And that's part of why this is so useful cuz if you're near an airport, you hear a lot of flights and you get a lot of signal. If you're not, then the silence

signal. If you're not, then the silence itself is also a signal. So it really works anywhere on that spectrum. So,

what that means for our search, I'll take all of these 73,000 points and I'll apply a silence filter. Feel free to pause here if you're interested in the specific thresholds and logic I used,

but overall I just disqualified any point that would have hurt a flight between our two events, and I was pretty conservative with the thresholds. So, to

recap, we took our velocity filtered flight segments for event one and event two, laid candidate points based on the slant distance range from the Doppler curve measurements. Then for all those

curve measurements. Then for all those candidate points did the time offset filter where that point would hear a flight of the right speed at the right distance away and then 920 seconds later

hear another flight of the right speed at the right distance away. That gave us this roughly 70,000 points. Now adding

on to all that logic the silence filter where that point doesn't hear another flight between the two events that drops the 70,000 to 15,000.

Even with all those constraints, there's just so much land area and so many flights that 15,000 points still meet all those criteria. And usually at this point in a search, I would just start

looking at some of the points, seeing why they don't work, and then encoding that logic as another filter. The

problem here is we just have this picture. There's nothing else we can get

picture. There's nothing else we can get from this. And we use pretty much

from this. And we use pretty much everything we could from the spectrogram. the velocity, the slant

spectrogram. the velocity, the slant distance, the time offsets, even the absence of the planes between the two.

There's just not much more we can do.

So, thanks for watching.

I'm just kidding. Of course, we're going to find it. Remember when I said this?

So, assuming linear constant velocity, there are only two things that determine what this Doppler curve looks like and this. What you'll notice on this one is

this. What you'll notice on this one is that it doesn't fit a perfect Doppler curve though.

The reason they don't fit an ideal Doppler curve is because planes don't fly in steady, perfectly linear motion.

They change heading, they speed up, they slow down. And all of that info is

slow down. And all of that info is encoded in this Doppler curve. But to

get it out, we're going to need to go a layer even deeper.

First, we looked at whether a flight occurred at this offset. Then, we looked at whether a flight of a certain speed occurred at this offset at a certain distance. Now what we're going to be

distance. Now what we're going to be asking is how exactly did that flight pass over? Now you can't go from

pass over? Now you can't go from spectrogram to knowing the exact flight geometry. Multiple things can cause

geometry. Multiple things can cause variations in this. It could be a change in speed, heading, altitude, any combination of those. So instead of trying to reconstruct what flight

geometry would have produced this, what we can do instead is model what the exact Doppler curve would have been for the actual flight geometry of each flight. We then do this for each

flight. We then do this for each candidate point on each flight path and then we find the one that matches.

So how do we model a Doppler curve?

The Doppler curve is a function of the radial velocity of the plane with respect to the observer. That sounds

complicated. All it means is how fast was the plane moving directly toward or away from the observer. So if this is us and we're right in front of the plane, plane velocity is 90, radial velocity is

90. If we're directly to the side, plane

90. If we're directly to the side, plane velocity is 90. Radial is basically zero. It's not going toward or away from

zero. It's not going toward or away from us. So at this point, there's no Doppler

us. So at this point, there's no Doppler effect. And anywhere in between, we can

effect. And anywhere in between, we can see that red line representing the radial component of the plane's velocity. You can see here how the

velocity. You can see here how the Doppler curve would change depending on the actual flight geometry. So how we'd actually create this Doppler curve. What

we want to know is what frequency did the observer hear over time. So we'll

start at t equals0 or cpa. In this

specific example from the flight data that's 71230. We don't know the actual

that's 71230. We don't know the actual time from the recording. This is just an example to illustrate. So for that time we go look at the flight data at 71230 and we get the plane's radial velocity.

In this case since it's at CPA the radial component is zero. But we then take that radial component and run it through the Doppler equation and it outputs a frequency. The Doppler

equation looks complicated. It's not

really. C is just a constant for the speed of sound. V subr, that's the radial component that we just calculated. The only other term in here

calculated. The only other term in here is F sub0. That's the fundamental frequency and that just comes from here where we measure the curve. Just

remember how we slide this up and down.

So that's all the components. It outputs

130 Hz. And then we plot that for 71230.

Then we just do the same thing again for 71235.

Check where the plane is. Get its radial velocity 62.7. run that through the

velocity 62.7. run that through the Doppler equation and get 109.9 and then plot that there. If you do that for all these points, you end up with the Doppler curve for this observer.

However, that's actually wrong and it would make our search fail entirely.

I'll give you a second to pause and think about why. The reason is we want to model the frequency heard by the observer at each of these times. So, the

frequency heard by the observer at 71230 is not based on the plane's radial velocity at 712:30.

It's actually based on the plane's velocity when it produced the sound that we then hear at 71230. So the correct flow is fort equals 0. Find where the plane was when it produced that sound.

Take that radial velocity, run it through the Doppler equation, and then plot that frequency. And so here's what that looks like. Now, if we just do that to the end, we're left with two curves.

One is the wrong way, one is the right way. And you might just think, okay,

way. And you might just think, okay, sound takes time to travel, so just delay the curve. What's the difference?

It's not just the same curve delayed in time. See, if I match up like this

time. See, if I match up like this point, very different curve. Same thing

over here. Match up that point. So, that

first methodology didn't just make the right curve, but a few seconds late. It

made an entirely wrong curve. Here's an

actual example. So, this is our specttogram. This is what that wrong

specttogram. This is what that wrong curve would have plotted versus this is the right methodology. So, that's how we model this Doppler curve for a given observer. We then just repeat this

observer. We then just repeat this process for all the candidate points and all of the flights. But what we're left with from that is the exact Doppler

curve that each point would occur based on real flight geometry. So then we have all these curves. The question is how do we score them? So we take the Doppler curve, we plot it on a spectrogram, and

then based on how close it looks, I raise up a scorecard kind of like gymnastic. I'm just kidding. We do plot

gymnastic. I'm just kidding. We do plot it on a spectrogram, but we just sum the brightness values that each curve covers. There's two things we have to

covers. There's two things we have to account for though. One is that the fundamental frequency that we used to model this curve was based on a measurement here. So that could

measurement here. So that could introduce some source of error. The

second is similarly the CPA time is as well. So the first thing we have this

well. So the first thing we have this modeled Doppler curve. How do we know where to put it on the spectrogram for scoring? Well, for the modeled curve, we

scoring? Well, for the modeled curve, we have a delayed CPA point. That's just

from flight geometry. But remember, we don't know the time of filming. So we

could note the delayed CPA time for this flight. But on our spectrogram, we have

flight. But on our spectrogram, we have no idea where that is. But we do have the CPA time in terms of just the time stamp of the audio. So that lets us say put the delayed CPA time right at this

time stamp. But back to what I was

time stamp. But back to what I was saying, that is a measurement and it's a source of error. So the way we account for this is when we place the curve on the spectrogram, we do a sweep of a few

seconds on both of those variables that we measured. So the slight time offset

we measured. So the slight time offset and then same thing for the fundamental frequency and then we just take the highest scoring combination of those.

That's how we score event one here.

Event two is a little bit different. We

still do the fundamental frequency sweep because that is still a measurement.

However, event 2's temporal placement relative to event one is entirely determined by the flight geometry.

Meaning if we correct for that error when we place event one and event two is still placed at a bad offset time that genuinely just means that this is not a good candidate and we want to capture

that too. So that's the combined locked

that too. So that's the combined locked scoring methodology where the offset between these two because it's determined only by flight geometry is entirely locked. We correct for

entirely locked. We correct for measurement error on event one and then event two falls wherever it does. That's

how we eliminate any potential measurement error, but still get really precise temporal offset scoring. The one

downside of this is that 200 meter grid resolution that we've been using is just a little bit too coarse for this. Like

the actual correct candidate could be up to 141 m off and like kind of score pretty bad. That's good because it means

pretty bad. That's good because it means it's really precise, but it also just means we have to take our 200 m grid points and expand them into 100 m resolution and just score more points.

All of this is done for one candidate point. So we need to do this for the

point. So we need to do this for the 15,000 times whatever to make them 100 meter resolution. So probably like 30

meter resolution. So probably like 30 40,000. So super computationally

40,000. So super computationally expensive and that's why you see all the steps we did prior to this to narrow down are still extremely important like you just could not do this on the full

US for the full day. So that's the entire methodology. I ran this for all

entire methodology. I ran this for all the points. Let's take a look at the

the points. Let's take a look at the results. So each candidate point has an

results. So each candidate point has an event one score and an event two score.

So I place them on this scatter plot here. Higher scores are better. So we're

here. Higher scores are better. So we're

looking for things in the top right quadrant. So there's some high scoring

quadrant. So there's some high scoring ones here, but there's not really one that stands out a ton. But for each combination of flight IDs, there's kind of a cluster of points. And really

within each cluster, what we care about is the highest scoring point. So I can do basically that on the scatter plot, group them by flight pair, and then just take the highest scoring point for each flight pair. And now you can see there

flight pair. And now you can see there is one that scores a lot higher. And

I'll show you the actual spectrograms and modeled curves for a couple of these top ones. So this high scoring point

top ones. So this high scoring point with 20.48.

That's this one here for event one. I

mean you can't even see the Doppler curve because it's perfectly covered.

And then for event two, very similar.

But let's also look at the second best scoring one. So you kind of see the

scoring one. So you kind of see the difference between a 20 versus a 15 score. So for event one, it's pretty

score. So for event one, it's pretty close. It misses this top bit right here

close. It misses this top bit right here as well as being a little bit off in the center. But I mean, as far as event one,

center. But I mean, as far as event one, this is a 95. This is a 10 six. What

really brings down the score is it's event two. And you can see it's event

event two. And you can see it's event two is pretty badly off. But this is the second best scoring point. And it's kind of similar for three, worse on event one, a little bit better on two. Like

none of these are even close. And then

first place is spoton. So let's go look at this point in CUGIS. So here are the scored points. Now that I know the

scored points. Now that I know the flight IDs though, what I can do is run much finer resolution with just that flight. So like here's the heat map when

flight. So like here's the heat map when you just score event one. And then for event two, you can see its flight path and its heat map. And then combined we have this.

But all this has just been dots and curves and scores. We still haven't found this location. I want to do one more thing before we hop into street view because this is probably the most

insane part of the entire project. Do

you remember this Doppler curve? the one

that you couldn't hear even when you were listening for it. What if you just had this one minute clip, but you knew the time that it was recorded? Could you

still find this? So, say I told you this was 8:37 p.m. UTC. We could do the same spectrogram measurement, get the velocity bounce for the flight, but then

also filter down to only the segments at that time. So, that's the view I have

that time. So, that's the view I have here. So, here are the scores for that.

here. So, here are the scores for that.

Since it's only one flight, there's no scatter plot. It's just a histogram, but

scatter plot. It's just a histogram, but right hand side is better. And here's

the view if you only take the highest scoring point per flight. So, we have a couple high scoring points here. And

here's the spectrograms. I did one thing slightly different on this, which is because this Doppler curve is so faint, I just kind of traced it with a line and then scored each candidate's curve based

on how close it was to that line rather than based on the actual brightness. So,

the top scoring point is spot on. Second

place is close, but I mean, you can definitely see where it's off. And

beyond that, it gets pretty bad. But

what you might notice about this coordinate, if we go back to our original two event scoring, the highest scoring point 40.87 40.8735 735

the same location scoring first by a significant margin across two different methodologies with entirely different flights. And if I were to generalize

flights. And if I were to generalize this, you can kind of pick two things here that are bad. So, in our first example, we didn't know what time the video was filmed or where in the US, but we had two flights. In the second

example, we only had one flight. We

still did the full US search, but we had a much tighter temporal constraint. So,

you kind of just need one of these things to be good. The thing is, though, I've never had a search where all I knew was that it was just somewhere in the US. You typically know a city or a state

US. You typically know a city or a state or even, oh, okay, I'm looking north at some body of water. Like I chose this picture just to demonstrate the extremes of like even knowing absolutely nothing,

you can still find the location. If you

knew the city, a city is like 01% of the land area. So when you constrain the

land area. So when you constrain the spatial bounds that much, it allows you to open up the temporal bounds and you don't need to know the minute or the hour or even the day of filming. I just

want you to grasp how powerful this is.

We got the location of that recording within a couple hundred feet from 1 minute of audio where you cannot hear the plane. And I mean, these flyovers

the plane. And I mean, these flyovers happen all the time. And those

frequencies go through walls. Like any

video you take could have that in there.

But anyway, we still have not found this exact photo. So, let's go back here in

exact photo. So, let's go back here in CUGIS. I'll zoom into kind of the center

CUGIS. I'll zoom into kind of the center of this. I mean, it's honestly in water,

of this. I mean, it's honestly in water, so it'll be like right on this bank.

Grab this coordinate and let's take a look in Google Maps and let's see what we got for street view.

All right, let's zoom in here. So, what

we're looking at is a super pixelated tree line, and we kind of want this V-shaped tree followed by a point, and then this big rounded part. So, from

this location, nothing really that way.

I'm not seeing any V-shaped trees here.

Here's kind of one, but we needed to go up into the right.

There's one more here where we have this Vshape with a point and then this rounded part. Here they are side by

rounded part. Here they are side by side. And if I zoom out and look at the

side. And if I zoom out and look at the top of this pole, here is the exact security camera that took this footage.

So, how do you protect yourself? Well,

the first thing people might think is a highp pass filter. There are two reasons that this doesn't work. First, the

frequencies of these Doppler curves are pretty much the same frequencies of the human voice. You'd be removing a good

human voice. You'd be removing a good chunk of what you actually want to record. Second, depending on how you

record. Second, depending on how you highpass, it might not even work anyway.

So, say this is my recording. I have the Doppler curve here. I'll apply the highp pass and cut off the Doppler curve. I'll

export this as an MP3. Open that MP3.

And the Doppler curve is completely gone. Now, let me just boost all the

gone. Now, let me just boost all the frequencies. And the curve is back.

frequencies. And the curve is back.

The next thing you might think, well, I would be talking during my video.

There'd be other loud things going on, so the mic would pick up on that, not background plane sounds. What if I told you for the past 15 seconds I've had a

JBL speaker across the room playing a Doppler curve?

So, highpass doesn't work. You can't

drown it out. One thing that works is just making cuts. That way, you can't use the time offset between these. But

remember when we found the location with just 1 minute of audio? If it's a large search area, you might need to know the pretty precise time of recording. And

you might think, when are you ever going to know the exact time of a recording?

How about any phone call you've ever had? But phone calls compress audio.

had? But phone calls compress audio.

They apply noise reduction. Surely a

Doppler curve won't survive that.

Right now, you're listening through another phone speaker under a bunch of squish follows. I'm in a different room

squish follows. I'm in a different room with that same JBL speaker playing Doppler curve. Just to verify there's no

Doppler curve. Just to verify there's no leakage, I'm going to go on mute.

So if the Doppler curve was going through the phone and only through the phone, what you should have seen there was a gap in the Doppler curve and then when I started talking again, it came back.

So highpass doesn't work. You can't

drown it out and it can survive phone compression. So what can you actually

compression. So what can you actually do? Like I said, you could make cuts in

do? Like I said, you could make cuts in your videos or apply some additional noise cancellation. But look, I'm not

noise cancellation. But look, I'm not trying to fearmonger with this video.

The odds that someone can and does actually do this is pretty slim. In

reality, there's probably a hundred easier ways to find you than this. The

one real use case I could see, I mean, the audio we used is just from some random security camera, but it could have just as well been footage from a investigation or a video online.

Something with literally zero visual clues could be located using this. And

maybe that makes you feel optimistic, maybe a bit uneasy, or maybe like me, you just find all this fascinating. That

when you hear a plane fly over, you're probably in the only place on Earth hearing that exact curve. And all the information needed to recreate that

exists. And 30 minutes ago, all of this

exists. And 30 minutes ago, all of this was just background noise.