Can a Seestar do real science?

A Seestar is a small smart telescope: a 30 mm or 50 mm lens, a color camera and a mount in one box, run from a phone app. It was made for pretty pictures. But every picture starts as a stack of short exposures, called subs, and each sub records how bright every star in the field was at one moment. Measure those brightnesses and positions carefully, and you have data.

A thin line beside a passage says how far its claim is proven: measured, implemented. Colored numbers carry the same mark. A passage with no line is stated. Measured numbers come from published pages on this site, each linked where it appears. Numbers from anyone else are stated, and say whose they are. See how claims are marked.

What a small telescope can measure

Five kinds of work suit a small aperture, because they need careful timing or careful comparison more than they need faint stars.

Kind of work	What it takes	What I measured
Variable stars	Many subs through a night, each with its time, and steady comparison stars on the same frames.	5,223 light curves of 869 known variables, from Seestar Collective images. (measured)
Eclipse timings	A run that covers the whole minimum, and a clock you trust.	385 times of minimum or maximum that passed a 5-sigma test, from the same images. (measured)
Exoplanet transits	A bright host star, an unbroken run of several hours with a baseline before and after, and many comparison stars.	Two transits caught: HD 189733 b and WASP-12 b. (measured)
Star positions	Sharp focus, a good plate solution, and a reference catalog to check against.	Night-average errors of 48 to 66 milliarcseconds against Gaia on my own nights, and 72 to 86 on another owner's telescope with the method locked in advance. (measured)
Occultation timing	Short video exposures and an outside time reference, because the Seestar has no GPS clock.	Nothing yet. I haven't tried one.

Variable stars and eclipse timings

Members of the Seestar Collective shared their nights, and I measured every known variable star that sat on their frames through a night: every star with a known period in VSX, the AAVSO's Variable Star Index, brighter than magnitude 14.5. As of 2026-10-13, that is 5,223 light curves of 869 known variable stars, from 463 observing sessions (203,232 frames) taken between 2024-06-08 and 2026-10-11 (measured). A light curve is a star's brightness plotted against time.

67 stars clearly vary on a single night, and 46 more likely do (measured). "Clearly varies" means that on at least one night the star's swing is at least 5 sigma and at least twice that of its comparison stars, which stay flat.
5 stars "held up when sessions from two separate groups of telescopes were checked against each other", in the page's own words: the two groups were chosen before anyone looked at the data, and each showed the same variation on its own (measured).
385 times of minimum or maximum passed a 5-sigma test (measured). Timings like these are what eclipsing-binary observers collect to track how a pair's orbit changes.
The periods and star types on the page come from VSX, a third-party catalog.

The qualifiers travel with those numbers. The brightness scale is untransformed, so a level can sit a few hundredths of a magnitude off a standard V. These are preliminary measurements, and I have submitted none of them to the AAVSO; the people who took the images can, under their own observer code (measured).

Data: the Seestar Collective. See every curve on Seestar Collective light curves, and look up your own telescope there. New to light curves? How to read a light curve.

Exoplanet transits

When a planet crosses in front of its star, the star dims by a percent or two for a few hours. A 30 mm lens can see that. I've caught two.

HD 189733 b, with a 30 mm Seestar (Experiment, 2 October 2026, Exploratory): the star dimmed by 2.50 ± 0.34 %, and the middle of the transit came 1.0 minute before the prediction (± 3.6 minutes) (measured). Large telescopes' published depth is about 2.45 % (third party). The reduction I froze before looking, green pixels only, gave 2.83 ± 0.40 % (measured). Catching HD 189733 b.
WASP-12 b, with a 50 mm Seestar (Experiment, 11 October 2026, Exploratory): a dip 1.63 ± 0.20 % deep, 8.3 times its own uncertainty, with mid-transit "within half a minute of the newest published prediction" (+0.4 ± 4.2 minutes) (measured). Published depths run from about 1.25 to 1.54 %, depending on the paper and the color of light (third party). ExoClock lists a 5-inch minimum aperture for this target (third party). In the Experiment's own words: "Exploratory: one night, one telescope." Catching WASP-12 b.

What mattered most on the HD 189733 b night was the comparison stars, the steady stars in the same frames that the target is divided by. Going from 7 hand-picked comparison stars to 98 chosen by fixed rules cut the scatter of the 10-minute averages 2.3 times and moved the mid-transit from 14 minutes late to on time (measured).

To try one yourself: Catching an exoplanet transit with a Seestar.

Star positions

Measuring where a star sits on the sky is called astrometry. I measured hundreds of star positions on two nights of my own Seestar S30 Pro frames of M 31 and M 33 (10-second subs, alt-azimuth mount) and checked each one against Gaia DR3, the European Space Agency's star catalog (Experiment, 18 September 2026, Confirmed):

Averaged over a night, a G 11-12 star's position is typically off from Gaia by 48 milliarcseconds on the M 31 night (171 stars) and 66 milliarcseconds on the M 33 night (133 stars), counting stars seen in at least 20 frames and not pulled by a close neighbor (measured). These two nights are the ones the method was built on, so the Experiment lists them as not confirmed. A milliarcsecond (mas) is a thousandth of an arcsecond; one pixel of this camera covers 3.67 arcseconds, so that is "under a fiftieth of a pixel".
I then locked the method and ran it once, unchanged, on frames from three other owners' S30 Pro telescopes, set aside unopened until that run. On one of them, with an IR-cut filter on an equatorial mount, it gave 72 and 86 mas on two nights (706 and 607 stars), inside the ranges I had predicted, and across the test it met 19 of the 22 predictions I had written down in advance (measured). Not all of the set-aside frames were untouched: 108 of that telescope's 617 had gone into an earlier test, and frames from the same two nights were used to rehearse the method; the Experiment records all of this. That is what is confirmed: the method, on telescopes it was not tuned on.
It doesn't always work that well. On another owner's telescope the night averages came out at 146 and 265 mas, and on my own transit night, through a light-pollution filter, 194 mas (measured). Sharp focus and stars that move around the sensor are what help most.

How precisely a 30 mm smart telescope places a star.

Hubble's Cepheid V1 in Andromeda

In 1923 Edwin Hubble found a star in Andromeda that brightens and fades every 31.4 days, and it showed that Andromeda is a separate galaxy. In frames almost all from the Seestar Collective, the Experiment's summary reads: "I located Hubble's famous Cepheid in Andromeda in smart-telescope frames. Its pulse is not confirmed yet." (Experiment, 6 October 2026, Exploratory.)

Located: in 28.4 hours of frames, V1 comes out brighter than all 2,185 empty patches of the galaxy's disk measured the same way, about 3.5 sigma once the patches' heavy tail is counted (measured). That is a solid measurement at a known position, not a discovery.
The pulse: two tests I wrote down in advance did not see the 31.4-day swing above the noise. A later look reaches about 3 sigma across all archive frames, which I had already looked at, so not enough to call it (measured).

Data: the Seestar Collective. Finding Hubble's Cepheid V1 in Andromeda.

What other owners report

These are other people's results, paraphrased. I haven't repeated them.

Brightness against other observers. On the AAVSO forum (third party), one Seestar S50 owner reports green-channel estimates of long-period variables that differ by 0.05 to 0.1 magnitude from other observers' green-filter or visual estimates. Another reports that eclipsing binaries with about 1 magnitude of variation look good after averaging 4 to 6 images at a time, but not at the 0.01 magnitude level from a city sky. A third reports that EQ mode has helped long time series, because field rotation is no longer an issue.
How to label it. Advice on the same thread (third party): report green-channel measurements as TG, using the comparison stars' known V magnitudes, which makes them untransformed photometry.
Occultation timing. An occultation is a star briefly blinking out as an asteroid passes in front of it, and its timing pins down the asteroid's size and shape. A report from IOTA/ES (version 2.0, December 2025; third party) tested a Seestar S50 for this. It found the raw video of the Solar System mode is the only usable recording, that the Seestar takes its time from the phone network with no GPS pulse, and that in simulated events the timing error stayed within one frame duration. It concludes the S50 can work for suitable events, as long as a flashing device driven by a GPS pulse is used as a time reference, and that practical experience with real occultations is still lacking.

Where to send data

Variable stars: the AAVSO, the American Association of Variable Star Observers, through its WebObs upload under your own observer code. Submitting to the AAVSO walks through it.
Exoplanet transits: ExoClock, a project that collects transit timings to keep exoplanet predictions current.
Occultations: IOTA, the International Occultation Timing Association, or its European section IOTA/ES.

What you need

Save the subs

A measurement is made from the individual subs, not the stacked picture, and the app saves them only if you switch that on before the session. What your Seestar saves, and which files to keep.
Use EQ mode for long runs

In alt-azimuth mode the field turns through the night, which smears stars near the edges. On an equatorial mount, on a wedge, the field stays put and the stars sit on the same pixels all night. When it's worth the setup: Alt-az or EQ mode: why the field turns and what it costs.
Trust your clock, and start early

Timings are only as good as the frame times. On the WASP-12 b night I used the telescope's own clock without checking it against an independent source, which is an assumption, not a measurement (measured). On HD 189733 b I started late and had about 20 minutes of steady data before the transit, which widened the timing error bar (measured). Next time I start at least an hour early.
Share a night

To help test and improve SkyCruncher's measurements, send a night of subs to SkyCruncher. Your observing site, the latitude and longitude in each file's header, is replaced in your browser by the center of the 6 km map cell it falls in before anything is sent (implemented). The files stay private; the target, its position on the sky, the number of frames and the hours of exposure appear as a dot on the public sky map (implemented). See Privacy.

Sources

Seestar Collective light curves: the variable-star totals, labels, timings and caveats (as of 2026-10-13; the page says it was measured in October 2026).
Catching HD 189733 b with a 30 mm smart telescope and Catching WASP-12 b with a 50 mm smart telescope: the transit depths, timings and lessons.
How precisely a 30 mm smart telescope places a star: the star-position errors and the test on other telescopes.
Finding Hubble's Cepheid V1 in Andromeda with smart telescopes: V1 located, pulse not confirmed.
AAVSO forum (third party): Using the ZWO Seestar S50 for photometry. Checked 2026-10-20.
IOTA/ES (third party): Seestar S50 for Occultation Work, version 2.0, December 2025. Checked 2026-10-20.