How to read an action camera spec sheet

The first fork is not brand but shape. A flagship action camera is a sealed, mountable body with a screen and the current stabilisation — the class GoPro, DJI and Insta360 actually compete in. Compact and wearable cameras trade size for everything else, though two of them now carry a flagship-sized sensor in about fifty grams. 360 cameras record the whole sphere and make you reframe it. Budget action cameras look similar on paper and differ underneath, most sharply on horizon lock and bare waterproofing. Pocket gimbals are not action cameras at all and are here because people buy them instead of one.

The catalogue currently holds 13 360, 13 budget action, 9 flagship action, 6 compact / mini and 1 pocket gimbal.

“1/2.3-inch” and “1/1.3-inch” are optical format names inherited from video camera tubes. The fraction described the outside diameter of a glass tube, not the imaging area, and the convention survived into digital sensors where it means nothing intuitive at all. Worse, it compresses exactly the difference that decides low-light performance: the labels look a third apart and the sensors are a factor of two and a half apart.

So every camera here carries a derived area instead, in square millimetres, from the standard table below. It runs from the SJCAM SJ11 Active at 18 mm² to the Ricoh THETA Z1 51GB at 116 mm² — a factor of 6.4. For a 360 camera the figure is the area of one lens’s sensor, because the noise in a reframed shot comes from the single lens that was pointing that way.

One honest caveat. DJI builds two of its sensors square rather than 4:3, and markets the square 1/1.1-inch chip as equivalent to a one-inch sensor. Worked off this table, a 1:1 crop of that image circle is about 96 mm², not a true one-inch’s 116. We publish the table figure and say so rather than repeating the marketing equivalence, and the square readout is a genuine advantage for a different reason — it gives a vertical crop as much room as a widescreen one.

Aperture is the other half of the low-light equation. Across the catalogue it runs from f/1.6 to f/2.85, which is more than a stop and a half, and it is the cheapest way a smaller manufacturer can compete with a bigger sensor.

Almost every camera here out-resolves what its buyer delivers. Resolution is worth attention for two reasons only: cropping room in the edit, and whether the camera can hold a useful frame rate at its top mode. A camera that shoots 8K at 20 frames a second — and one here does — has an 8K stills mode with a video label on it, because 20 fps is below the rate that reads as motion.

The trap is comparing a 360 camera’s resolution to a flat camera’s. They are not the same unit, and the next section is about why.

A 360 camera records its full width across a complete circle. When you reframe a normal 16:9 clip out of it, only the slice your field of view covers is used. The arithmetic is:

usable width ≈ sphere width × (field of view ÷ 360)

Take the GoPro MAX2, one of 7 cameras here tied at the top spherical width. It records 7,680 px around. At a typical 80° reframe that is 7,680 × (80 ÷ 360) = 1,707 px — roughly 1.7K, which is under 2K.

That is not a criticism of 360 cameras; it is the price of being able to choose the angle afterwards, and the reframed footage is still perfectly good on a phone. It is a criticism of comparing the number on a 360 box to the number on a flat camera’s. This site shows the estimate on every 360 camera’s card, detail page and compare table, labelled as an estimate — stitching, supersampling and the projection move it a few per cent either way.

The other thing the spec sheet does not tell you is that every clip needs a reframing pass before it is watchable at all. That editing time, not the resolution, is what usually decides whether a 360 camera keeps getting used. Whether a 360 camera is worth it covers that in full.

Resolution sets how many pixels there are; bitrate sets how much information is kept about them. It is what decides whether spray, leaves, snow and anything else with fine moving texture holds together or turns to mush. Across this catalogue the published ceilings run from 25 to 360 Mbps, a factor of 14, which is a far wider spread than the resolution numbers suggest.

Two cautions. Makers quote the ceiling from whichever mode reaches it, so a headline figure may belong to a timelapse rather than to normal recording — that is why this site labels the row as the published ceiling. And 18 of the 42 cameras here publish no bitrate at all, almost all of them in the budget tier. Bitrate is also what decides your microSD card: a card that cannot sustain the write rate is the most common cause of a recording that fails without explaining itself.

Thirty frames a second is the floor for anything that moves. Sixty is where panning stops strobing. Anything above that is slow motion: 120 fps played at 30 is four times slower, 240 is eight times. What matters is the highest frame rate at the resolution you actually deliver, which is why this site records the 4K ceiling and the 1080p ceiling separately rather than quoting the camera’s best number at its worst resolution.

The 4K ceiling is the most useful single figure, and here it ranges from 30 to 240 fps. A camera capped at 4K30 cannot give you slow motion at delivery resolution at all, which is the limitation that separates most of the budget tier from the flagships far more sharply than image quality does.

Almost every camera here stabilises digitally: it reads a larger frame than it outputs and shifts the image around inside that margin to cancel shake. HyperSmooth, RockSteady and FlowState are three brands’ versions of the same idea, and all of them cost field of view and some resolution to work. Under hard vibration the weaker implementations warp straight lines, which owners call jello.

The exception is the pocket gimbal class, which holds the camera still mechanically on three motorised axes. That costs nothing in resolution or field of view and produces a distinctly different floating look, at the price of an exposed moving assembly that is the first thing to break in a fall.

The budget tier is where this specification is most oversold. Electronic stabilisation appears on almost every box in the category, and owners of the cheapest cameras consistently describe it as cropping the frame without meaningfully steadying it.

Stabilisation removes shake; horizon lock removes rotation. They are different settings solving different problems, and a camera with excellent stabilisation and no horizon lock will hand you a beautifully smooth shot of a tilted world every time the mount twists.

Full 360-degree lock holds the frame level through any rotation, including a complete roll. The 45-degree kind corrects lean within that range and gives up beyond it, which is fine for walking and not for a motorcycle through a corner. Of the cameras here, 25 have the full version, 2 the partial one, and 15 have none at all — almost all of the last group in the budget tier.

It costs field of view, because the camera needs image outside the frame to rotate into. That is one reason the tall 8:7 and square sensor readouts matter: a squarer capture has more to give the lock before the crop becomes obvious. 360 cameras get it free, because they already have the whole sphere.

Eight-bit video records 256 levels per colour channel; ten-bit records 1,024. The difference is invisible straight out of the camera and very visible once you push the grade, because eight-bit footage bands in skies and gradients as soon as it is stretched. A log profile is separate: a deliberately flat curve that keeps highlight and shadow detail the standard profile would clip, which looks washed out until it is graded.

Both matter only if you grade. If footage goes straight from the camera to a social feed, ten-bit and log cost storage and buy nothing. If you colour-correct, they are the difference between recoverable and not.

One finding worth stating plainly: 24 of the 42 cameras here do not publish a colour bit depth at all. That includes most of one major brand’s line, where only the newest model states it. Where this site shows “not published” in that row, it is because the manufacturer does not say — not because we did not look.

A camera that needs its housing to get wet is a different object from one that does not. The case fogs, the seal is one more thing to check before every session, the audio goes muffled, and the air gap makes the camera run hotter. This is also where the budget tier is least straightforward: a box advertising 40 metres usually means the plastic housing inside it, and the camera itself has no rating at all.

Bare ratings here run from 1 m on the AKASO Brave 7 LE to 20 m on the Insta360 X6. 7 cameras are waterproof only in their case, and 4 have no water rating of any kind.

Two practical notes from owners. Flooding almost always happens through the battery-door gasket rather than a crack, so check it is clean and seated every time and rinse the camera in fresh water after salt or chlorine. And at least one major manufacturer’s warranty explicitly does not cover water damage, which makes the door seal your responsibility rather than theirs.

Every maker quotes runtime at the least demanding setting it has. Most use 1080p at 24 frames a second with the radios switched off, which is a mode almost nobody shoots in. The split-body cameras quote the figure with their battery handle or dock attached, which doubles or triples it — the camera module alone runs a fraction of the advertised time.

So a runtime number without its setting is marketing, and this site never shows one without the other. The claims here range from 50 min to 7 h 17 min, and the settings behind them are not comparable at all.

Where a named outlet has run its own test we print that too, with its conditions. Currently 4 cameras carry one, all four from a single group test in which the cameras recorded simultaneously until they shut down. That is a small number, and it is small because a same-setting comparative test is rare. We would rather show four real figures than forty estimated ones.

Cold is the other half. Most of the cameras here are rated to −20 °C; one is rated only to 0 °C, which is a genuine limit on a ski trip. Cold takes capacity out of any battery regardless of the rating, so winter shooting means spares kept warm in an inside pocket. Heat works the other way: a camera that repeatedly runs to a thermal shutdown degrades faster, which is covered in why action cameras overheat.

The lens on an action camera sits proud of the body, and on a 360 camera it bulges well past it, because a fisheye has to see beyond 180 degrees to stitch. That makes the glass the first thing to touch anything, and a scratch in the middle of a lens is in every frame from then on. It is the most common reason a working camera gets retired.

13 of the 42 cameras here sell their front glass as a user-replaceable part, which turns a write-off into a cheap swap. That change arrived recently and owners treat it as the most useful thing any 360 camera maker has done in years. Screw-on lens guards are the cheaper insurance and are close to universal among long-term owners — they cost a little contrast and some flare risk, which is covered in do lens guards hurt image quality.

Two systems dominate. GoPro’s folding fingers have been unchanged since 2014, which means a decade of clamps, harnesses, helmet mounts, suction cups and third-party rigs fits any camera with them. DJI and Insta360 use magnetic quick-releases, which are genuinely faster with gloves on and need an adapter to reach that ecosystem. Several cameras carry both, and a 1/4-20 tripod thread appears on most of the 360 cameras and the pocket gimbals.

This matters more than it reads. Someone who already owns a chest harness, a handlebar clamp and three helmet mounts is not choosing between two cameras; they are choosing between one camera and one camera plus a box of adapters. It is the most common reason people stay with a brand whose spec sheet they do not prefer.

Two of the three major brands sell an annual plan and gate features behind it: cloud upload, unlimited storage and camera replacement in one case, cloud storage and a replacement scheme in the other. One of them also holds its best stabilisation mode behind a higher tier, which is a capability of the camera you already bought being sold back to you. The third brand sells no plan at all.

That makes a straight price comparison misleading in the first year, so every camera here carries a year-one figure: price plus the plan, counted only where the plan gates something the camera advertises. It is an imperfect measure — some of what a plan gates is genuinely optional — but it stops a brand that charges separately for its features from looking cheaper than one that does not. What the plans actually gate has the detail, including the upgrade credit that turns out not to be one.

Every price here is a hand-maintained snapshot from the last catalogue update, marked approximate because action camera pricing moves hard. Launch pricing, clearance on the outgoing generation and seasonal promotions routinely move the same camera by a third, and at least one camera in this catalogue was discounted by hundreds of dollars within weeks of release. The retailer’s price at checkout is the real one.

This site carries no retailer links of any kind, no affiliate tags, and nothing that earns a commission. How we compare says where every number comes from and what this site refuses to do.