Frequently asked.

Who wrote UFO@home?#

It is developed by RR0, which among other things built rr0.org, a French encyclopaedia of unexplained phenomena. That site uses UFO@home to illustrate its own case files — and RR0 has taken care to keep the tool usable by anybody.

Where does the name come from?#

From SETI@home. In its day that application ran as a screensaver, so that anybody could lend their computer to the calculations of the search for extraterrestrial intelligence, and perhaps find on their own machine an interesting signal from a potential extraterrestrial intelligence. UFO@home borrows the idea: the work is done at home, on your own machine, in your browser, and anybody can take part in it.

What does the licence allow?#

UFO@home is under the MIT licence. That allows you to:

• use it on any site, including a commercial one;
• modify it, and keep your modifications private;
• redistribute it, bundle it into your own product, and charge for that product;
• fork it, and continue it in your own direction.

The one obligation is to keep the copyright notice and the licence text with the copies you distribute. There is no contributor licence agreement and no registration.

The source is on GitHub, including the scripts that generate its star, comet, nova and satellite catalogues from public sources — so the data is reproducible as well as readable.

What does it send over the network?#

Recordings stay in your browser: there is no account and no server-side storage. Save writes a file to your disk, and that file is the whole recording. What is kept, what third parties receive, and who answers for what is published: see Privacy.

Everything that is fetched — from where, when, and under which licence — is listed source by source on Sources and choices, the one list there is, so that this answer does not have to keep a second copy of it. In short:

• While you author, the editor looks things up for you: a place's coordinates when you press Locate, the weather record once a full date and a place are known, the relief and aerial imagery around the observer. Each lookup is a picker in the interface, where its data is reported, with the attribution its licence requires. Drafting a recording from its description sends that text to the model you chose, with your own key, and only when you ask.
• A page that only replays a finished recording (<rr0-scene>, <rr0-sighting>) asks for the relief and imagery tiles, what this site serves itself (satellite elements, 3D models), and what the recording itself points at (a photo of the place, a sound). A published recording carries its own weather and is never looked up again, which is also why it reads identically years later.

Can I host it entirely myself, with no external service?#

Yes, and nothing about it depends on RR0. The components are published on npm as @rr0/ufoathome and served from this domain; install them or copy the built .mjs bundles onto your own server and the reconstructions will play, with no runtime dependency on rr0.org. Recordings are your own files, on your own host, and a reconstruction asserts what its recording states and nothing more — no conclusion, no branding, no case number other than one you put there yourself.

Copy the .mjs bundles onto your server and the reconstructions will play. Skip the terrain and imagery providers (or point them at your own tile server) and the scene falls back to a plain horizon. The lookups exist for authoring; playback of a finished recording does not need them.

Why build this at all?#

The idea came out of conversations with the sociologist Pierre Lagrange, who would cite Roger Shepard's work to its author.

Because an account written down loses almost everything about it. In 1968, at the AAAS symposium, the psychologist Roger Shepard argued, in the paper he gave there, that a visual reconstruction of an account is more faithful than a written or spoken one — observers are much better at recognising and adjusting a picture than at generating a description. UFO@home is that idea, built as software: draw it, move it, and let the observer correct it until it matches.

The project began in 2003 as a Java applet. It was rewritten from scratch in TypeScript in 2026, as web components, because the applet had become unrunnable and because the missing half had always been the sky: a reconstruction without the real sky of that night cannot be checked against anything.

Why is an account drawn flat, and when does 3D come in?#

The account itself is drawn as it reached the observer's eye: a shape in their field of view, this big, moving this way. Because a 3D object placed in the scene is already a conclusion. It asserts a size, a distance and a solidity that no observer could perceive — and, worse, it silently rules out every explanation in which there was no object there at all: a halo, a planet, a satellite, an aircraft's landing light, a lenticular cloud, a reflection on a windscreen. Those links go to rr0.org's own catalogue of misperceptions, which is the reason this tool is built the way it is.

A flat shape on the observer's own field of view asserts exactly what they claimed: this is what reached my eye, this big, moving this way. Everything else stays open — which is the only way a reconstruction can be used to test a misperception rather than to rule one out by construction.

3D comes in as an interpretation, kept apart from the account: what the observer believes they saw (at Silly-le-Long, T1 describes one flat grey triangle, seen from the front and then from behind), or an analyst's explanation stated in a case file (there, the GEIPAN's light aircraft on final). Each is one choice of the player's Interpretation selector, shown one at a time, with a model, a size and a track in metres. The Compare with the account button then lays the account's own outlines over it and states, instant by instant, how far apart they are in direction, width and height. The account never changes to fit an interpretation: the interpretation answers to it.

It also means the tool is honest about size. A recording stores an angle, never metres — but it does NOT mean distance is given up on. Where the observer saw the phenomenon cross something whose position is known, that crossing is stated in the recording (it passed behind that hangar, in front of that tree) and the scene raycasts the exact line of sight to measure it. Each crossing bounds its real width from one side, an angle plus a distance is a size, and a size does not change as it moves — so every other instant of the recording reads back as a distance too. The editor prints the range under the apparent size, reports a contradiction where the stated crossings cannot all be true, and says “unknown” where nothing crosses the line of sight at all, which is the honest answer for a light in an empty sky.

Is a reconstruction evidence?#

No. It is a way of stating an account precisely enough that it can be laid beside the record — and the record is what does the work: the Moon's phase that night, whether the sky was overcast, whether a comet was up, whether low orbit was even sunlit.

What a reconstruction is trying to do is agree with the observer. Not to illustrate their account, and not to correct it: to converge on it, one correction at a time, until they say that is what they saw. Every step of that convergence adds something that can be checked, so the closer it gets, the more there is to explain the sighting with — or to fail to.

And that is where it earns its keep. Suppose the reconstruction can put a known phenomenon — a planet, a halo, an aircraft — in the exact place the observer saw something, at the right size, at the right hour. Show it to them and ask the only question that matters: does that look like what you saw? If the answer is yes, investigator and observer have converged on the facts rather than argued about the conclusion, and the explanation stands on something firmer than anyone's opinion of the other.

If the answer is no, that is a result too, and a better one than a written note saying the observer disagreed — because the thing they are rejecting is on screen, at the right place and the right size, and the next person can look at it.

Does it replace the investigation?#

No: it rests on it. A recording is a simulation of the observation, built from what the investigation established: the account, the place and time, the weather, the measurements. Laid beside the investigation, it can help check that an account is plausible, verify a measurement, show what another observer or another point of view would have seen, bring out the gaps between accounts or between an account and the record, and test a hypothesis (do a planet, an aircraft, a flock of birds fit the angles that were given?). Or it can simply illustrate the investigation and summarise it visually. Investigating stays the investigators' work.

How is this different from Sitrec?#

Sitrec (“situation recreation”, Mick West, from 2022) is an excellent tool and it does a different job. It starts from instrument data — a video, an ADS-B track, a sensor's own metadata, orbital elements — and reconstructs the geometry that could have produced that footage. It is the right tool when there is footage.

UFO@home starts where there is none: a person, an account, and a date. It reconstructs what an observer described, and it is built around what an account can and cannot say — angles rather than metres, a stated appearance rather than a placed object, and an explicit record of who supplied every non-testimonial fact.

Their licences differ too. Sitrec was MIT-licensed; it was archived in March 2026 in favour of Sitrec2, whose licence permits personal or non-commercial academic use only, and neither redistribution nor modification. UFO@home is MIT.

Isn't this what SIMOVNI did?#

It is the same idea, fifty years earlier and in hardware. SIMOVNI was an optical simulator built in 1976 within France's GEPAN (the CNES's own UFO study group) by Jean-Jacques Velasco: the observer looked through a viewfinder at the real landscape with a virtual UFO superimposed, whose shape (on slides), colour, brightness and size could be adjusted until it matched. It is, as far as we know, the first serious attempt at visual rather than written account. It was also a single physical bench in Toulouse, it was used on only one or two occasions, and Dominique Caudron — who had built his own simulators earlier — thought its results unreliable.

SimOvni 2, presented by Laurent Chabin at CAIPAN 2 in 2022, revives the idea with a head-mounted display and real calibration work behind it.

What UFO@home adds is not a better optical bench: it is that a reconstruction becomes a small file anyone can open, replay, embed, check and disagree with — instead of a session that happened once in a room, in front of one investigator, and left a written summary.

Why not just use Stellarium?#

Stellarium is a planetarium, and a very good one — we have read its source to learn how it does things. But it answers “what was in the sky?”, not “what did this person see, and could the sky account for it?”. It has no account format, no observered phenomenon, no weather, no decor around the observer, no long-exposure instrument, and nothing to embed in a page.

Where the two overlap, UFO@home sometimes takes the harder road on purpose: the Milky Way is a texture in Stellarium and a line-of-sight integral here, which is why its dark rift comes out of a dust model rather than out of an image. Where Stellarium renders something we do not, the working rule of this project is to go and find out how they do it.

Why not ask an AI to build the simulation from scratch?#

Because what it built would be an illustration. Asked for “a scene of what this witness saw”, a model writes a 3D scene or paints a video that looks right: a sky, a field, a light. Whether the Moon was really up, where the Sun really was, what the cloud cover was that morning, how big the thing really looked from where the observer stood: each of those is something it may have got right, and nothing in the result says which. It is also a one-off: a thousand lines of code nobody has checked, written differently for the next case, which cannot be laid beside the next case.

A UFO@home recording splits the work the other way. What the AI writes is only the account: a date, a place, directions and apparent sizes over time, each value marked as stated, worked out or assumed. The sky, the weather records, the ground, the optics of the eye or the camera are computed from that by the same engine for every case, tested, and public. The format's limits are part of the point: it has no field for a real size or a distance, so an assistant cannot slip a conclusion in among the facts, and what it does write is a short file anyone can read, replay, correct, and show the observer.

That is also why the two are not in competition. An AI assistant is a good way to write a recording from a case file: on a GEIPAN case, one given only the public file and these pages produced a recording whose phenomenon matched the one made by hand to within a tenth of a degree. What it cannot do on its own is make the result checkable, and that is the part this project is.

How do I ask for a feature, or report something wrong?#

Open an issue: github.com/RR0/UfoAtHome/issues. A GitHub account is free and takes a minute.

What makes a request easy to act on:

• A real case, if there is one. “Observers often say it was behind a hill” is a feature; “here is a sighting where that matters, dated, placed and sourced” is a feature that gets built right.
• What the record says. If the thing you want reconstructed is measured somewhere — a dataset, a catalogue, a published measurement — say where. This project's rule is that nothing gets invented: something reproducible needs data or physics behind it, and knowing which one it is settles most of the design.
• What you would conclude from it. A feature that cannot change anyone's reading of a case is decoration.

Pull requests are welcome on the same terms as any MIT project. If you would rather not use GitHub, the maintainer's contact is on rr0.org.

Can I put a reconstruction in a forum post?#

If the forum allows raw HTML and a module script, yes — the two lines are on the editor page, and every published reconstruction hands them out itself. Most forums do not allow that, for good reasons. Two things that usually work instead: an <iframe> pointing at a page of your own that holds the component, or simply a link to ufoathome.org/edit/?file= followed by the URL of your recording, which opens it here for anyone.

What if I don't know the exact date, or the exact time?#

Say so, and the tool will say so too. Dates are stored in EDTF, so “1954”, “June 2025”, “around 05:00” and “uncertain” are all things the format can state. What you lose is only what genuinely depends on the missing part: with no full date and place there is no sky to compute, and the interface says that rather than drawing a plausible one.

What languages does it speak?#

The components and this site are in English, French, Spanish and Italian, picked from your browser's own preferences with English as the fallback. Adding a language to the components means adding one typed messages module per component — no markup changes, no build configuration. It is one of the easiest contributions to make.

There is no language picker, and that is the point: an address carries no language, so a link you send is read by whoever opens it in theirs. Send someone in Lyon a section of this page and they arrive at that same section, in French.

Can I install it as an app, and use it offline?#

Yes. UFO@home is a web app you can install: in Chrome or Edge use the install button of the address bar, on an iPhone or iPad choose Share then Add to Home Screen, on Android choose Install app in the browser menu. It then opens in its own window, with its own icon.

Offline, the home page, the player and the component are kept on your device, and so is every page, recording, model and map tile you have opened since — enough to replay an account again in the field with no signal. What needs the network stays out of reach without it: the weather and place names fetched for a new recording, imagery, and any recording you have never opened.

Online, nothing changes: the site always asks the network first, so you read the current version, and the copy on your device is only used when the network does not answer. Removing the app, or clearing the site's data in your browser, erases that copy.

What does it cost?#

Nothing: no price, no account, no quota and no paid tier. The external services it uses while authoring are public ones whose usage policies it respects — which is why place search runs only when you ask for it.