The 112 UAP Videos That Can't Prove Anything—and Why Physics Is to Blame

In 1996, an infrared sensor aboard a U.S. military aircraft captured something crossing its field of view. The object traversed the frame in just a tenth of a second, moving at 142 pixels per frame. It looked fast. It looked anomalous. It could also, according to a new scientific analysis, have been a bird flying past the camera lens. The same pixel movement fits equally well with a craft traveling at Mach 5 a kilometer away. And there's no way to tell which is true—not from the video alone.

This is the core finding of a paper by Jacob Haqq-Misra of Blue Marble Space and Ravi Kopparapu of NASA Goddard, who examined all 112 sensor videos released under the PURSUE (Presidential Unsealing and Reporting System for UAP Encounters) initiative. Their verdict is blunt: without additional data, these videos cannot resolve whether any of the objects they capture move anomalously fast. They can neither prove exotic behavior nor rule it out.

The Science

The PURSUE initiative, launched by the U.S. Department of War between late 2025 and mid-2026, released five tranches of UAP records—112 sensor videos from targeting pods and turret-mounted cameras on military aircraft, including both infrared and electro-optical imagery. Haqq-Misra and Kopparapu set out to determine what these videos can actually tell us about the velocity of the objects they capture.

The mathematics is straightforward in concept but demanding in practice. Any video gives you the rate at which an object's image moves across the frame—pixels per frame. You also know the frame rate and frame size from the video file itself. But converting pixel movement into physical velocity requires four additional pieces of information that the researchers call out by name: the range to the object, the velocity of the aircraft doing the observing, the aspect angle between the object's path and the viewing direction, and the field angle of the camera sensor. None of the 112 videos provides this complete set.

The analysis builds on what's called the "velocity equation," which relates the angular rate of an object's line of sight to its physical motion relative to the observer. The key insight is that a single camera records only angles—it sees things sweep across its field of view, but it cannot directly measure how far away those things are or how fast they're approaching or receding. This is why multi-sensor observation, the kind being developed by projects like the Galileo Project at Harvard and the UAPx team, is the only way to truly constrain an object's kinematics. A single camera leaves fundamental ambiguities that no amount of analytical effort can overcome.

What They Found

The researchers classified all 112 videos by their kinematic behavior. Transit videos show objects crossing the field of view; tracked videos show objects the sensor has locked onto and follows; some videos contain both types of footage.

Kinematic Classification of 112 PURSUE Sensor Videos

Label	Value
Transit	43
Tracked	34
Both	35

Source: Haqq-Misra & Kopparapu, 2026. Classification based on kinematic properties of objects in each video.

Among the corpus, 43 videos capture transit events, 34 show tracked objects, and 35 include both. The distribution matters because tracked objects—which stay centered in the frame as the camera slews to follow them—have effectively zero pixel movement by design. This tells you nothing about their actual velocity. Only transit events, where something moves against the background scene, allow any pixel-rate measurement at all.

Even for transit events, the full velocity equation cannot be solved. The researchers examined every video for the eight quantities that appear in the governing equations. Frame rate and frame size are present in every file. But the angular scale of the camera, the range to the object, the observer's velocity, and the aspect angle are either redacted or absent from all 112 videos. "No video contains complete information to constrain the relative velocity," the authors state flatly.

Two clips received detailed individual analysis because they offered more than the others.

The first, DOW-UAP-PR113, is unique in the corpus because it contains visible depression-angle markings—burned-in graticule lines on the display—from which the camera's field of view can be reconstructed. Using these markings, the researchers calculated the camera's angular scale as approximately 2,030 pixels per radian, corresponding to a 54-degree horizontal field of view. The object in question crossed the frame at 142 pixels per frame, corresponding to an angular rate of about 120 degrees per second. This is the only clip in which this angular rate is constrained.

But range remains unknown. The researchers plotted the object's implied relative velocity against possible ranges and aspect angles, and the results span orders of magnitude. At very close range, the object could be moving at tens of meters per second—subsonic speeds consistent with a bird. At ranges beyond a few hundred meters, the same pixel movement implies supersonic or even hypersonic velocities. The object's size offers no rescue: its roughly 25-pixel image extent could represent anything from a small bird at close range to a 37-meter craft at 3 kilometers.

The second analyzed clip, DOW-UAP-PR149, takes a different approach. Rather than trying to recover the angular scale, it uses an object of known size visible in the frame as a reference: a carrier ship, approximately 150–200 meters in length, whose hull spans 920 pixels in the image. With this scaling reference, the angular scale cancels out of the equations, yielding a relative velocity in terms of the unknown range ratio between the object and the ship.

The result is an upper bound. If the object and ship are at the same distance from the camera, the object's relative velocity is no more than 99–132 meters per second—approximately 0.3 to 0.4 Mach. The researchers can rule out supersonic relative velocities for this particular case. But they cannot determine whether the object is at the same range as the ship or much farther away, which would change the calculation.

Velocity Upper Bound from DOW-UAP-PR149

Label	Value
Upper bound (same range as ship)	132 m/s
Mach equivalents	0.4 m/s

Source: Haqq-Misra & Kopparapu, 2026. Upper bound calculated assuming object and ship at equal range from sensor.

Why This Changes Things

This analysis matters because it sets realistic expectations for what government UAP disclosures can and cannot accomplish. The PURSUE release was substantial—112 videos from military sensors, representing years of collection. But the paper demonstrates that releasing sensor imagery without accompanying telemetry is like handing someone a stopwatch and asking them to time a race, but only giving them a blurry photo of the finish line. You can see that something happened. You cannot say what happened or how fast.

The parallel to the Navy's "GOFAST" video is instructive. That footage, released years ago, was initially interpreted as showing an object moving at extreme speed near the ocean surface. But when researchers obtained the on-screen telemetry and performed trigonometric reduction, they found the apparent motion was dominated by parallax—the camera's own motion as the aircraft banked over the ocean. The object's implied speed was consistent with wind-borne drift. The key was having enough information to account for the observer's motion.

The PURSUE videos lack even that much. The redactions remove whatever on-screen displays originally showed azimuth, elevation, and range information. The result is a dataset that may contain extraordinary objects or entirely prosaic ones—and no way to know from the footage alone.

This is not a failure of analysis; it is a demonstration of physics. A single camera records angles. Angles alone cannot close a kinematic problem. This limitation was identified decades ago by physicist James McDonald, and the PURSUE corpus illustrates it with perfect clarity.

What's Next

The paper does not argue that the PURSUE videos are worthless. It suggests three categories of data that would enable better constraints: flight logs giving the platform's velocity, sensor metadata describing field-of-view and line-of-sight pointing angles, and range records or correlated radar tracks. Not all of these will exist for every encounter—some may be passive flybys with no supporting instrumentation. But releasing whatever does exist for each case would help determine whether the objects exhibit anomalous motion.

The paper also notes that the videos may retain value for other purposes: studying the morphology of reported objects, understanding sensor artifacts and phenomenology, and sharpening the data requests that investigators should make for future encounters.

What the PURSUE release cannot do, according to this analysis, is settle the question of whether any of these objects move in ways that defy conventional explanation. The videos may show nothing unusual. They may show something extraordinary. The data released so far cannot distinguish between these possibilities. Resolving that distinction will require more than footage—it will require the supporting context that would allow physicists to actually solve the velocity equation rather than simply acknowledging its unknowns.