Starfield Navigator

From: Sol
To: Tau Bootis
↓ Scroll down for explanation ↓
(If there is no 3D map loading in this place, refresh the page. It's a bit buggy)


On 2022-06-12, the first gameplay footage of the upcoming space RPG Starfield was released. It revealed that the game features "over 100 star systems" alongside with this ~6 second video of the galaxy map of the game:

About ~75 stars systems were visible, some of them with names. Among those, three star names were "real": Sol, Alpha Centauri and Porrima. That implied that the world of Starfield is based on and located in the Milky Way. This project is the attempt to reconstruct the ingame galaxy map from that video.

Stars can be approximated as not moving, at least within the 6 second timeframe. Therefore, we used the pixel tracking feature of 3D programs (Cinema 4D in this case) to track the movement of the stars during that scene.

The program can then attempt to reconstruct 3D positions. Those are a best guess to their relation to each other, but can't possibly be correct in regard to their absolute position in the game.

Star positions can be represented in galactic coordinates, which are cartesian coordinates in a special frame of reference. A linear transformation from the arbitrary reconstructed coordinates to galactic coordinates was the key to advance the process. Luckily, in three dimensions we need exactly three positions to perfectly construct such a transformation. We used the positions of the three known stars to do so.

Next up was the process to identify all the other unlabelled systems. Databases of all known stars are readily available and were used to choose the closest match for each star. It was apparent that in most cases, that was unambiguous - the second-best match being a much worse fit. With that process, almost all stars were identified.

With the stars identified, the original constraint of Sol, Alpha Centauri and Porrima being exact matches was no longer required as we can just use common optimization algorithms to find a more general transformation to minimize the position deviations on all stars. A last optimization was to use that error as a metric to calibrate parameters for the original position reconstruction in Cinema 4D, in particular the assumed camera FOV. With all that, the position errors were less than 0.1 LY among a distance of ~100 LY. That puts a high confidence on the reconstruction.

On 2023-03-08, Bethesda Softworks released more video footage. It contained two additional names of real stars: Tau Ceti and Cassiopeiae. Tau Ceti is unambiguous and is positioned inside in the cluster of stars previously discovered, further confirming our method. Cassiopeia is a constellation of many stars, spanning among a considerable distance. It's unknown which one the game refers to, therefore a few of the brightest were added to this program. Note that only one of them is correct.

It's known that the game will use a jump mechanic to traverse between star systems. The video footage showed ships with jump ranges between 15 LY and 30 LY. With a galaxy map spanning at least 100 LY, multiple jumps will be required to move to farther points in the map. Choosing an optimal path is solved by Dijkstra's algorithm. This site has a routing functionality that plots such an optimal path between two points. If the final game requires that manual plotting is yet unknown. Other interesting facts learned:

The original starfield navigator is a C++ program written as a tool for the reconstruction process. This web version is more accessible and will probably switch to the ingame map once the game is released. It might transition into a companion tool depending on how details pan out. Both tools are open-source and the repository located at

Starfield Direct (2023-06-11)

This event featured both a new story trailer as well as the gameplay deep dive. In the new footage, about 40 location names could be seen. Most of them in the lower left location indicator of the HUD. Most of them were fictional:

Many of them are references to people or famous fiction. There were also a number of actual Stars:

For those we can look up their position in the usual star catalogues. But we can go a bit further with the fictional ones, too. There is a shot at 6:10 where the camera pans through the starmap. There, we can see Kryx, Olympus, Lunara among with some stars that were already positioned in earlier efforts. By tracking the stars in that sequence as before and aligning them to the previously reconstructed stars, we were able to identify them.

Note that we had some trouble with this tracking shot and this information isn't as high-confidence as the others. But we think that:

When we look at the position of all those new stars in relation to the previously recovered starmap, they fit right in. Interestingly they don't break the pattern of being "right" of Sol. And with stars like Algorab being 87Ly from earth, it seems increasingly likely that the often-quoted 50 ly radius might not be centered on earth. This is a look of the new systems (yellow) on the map. Note that right now, the navigator itself isn't yet updated.

We also saw many previously known Systems. Also some which were identified by this effort and now being confirmed, in particular Altair and Procyon.

It was also shown that an ideal path btween two systems with intermediate jumps is calculated by the game automatically, so no manual pathing is necessary. Therefore this functionality of the navigator won't be applicable in the game.