High fps burst culling: 20 to 120 frames a second

Electronic shutters made 20, 30, 60 and 120 frames a second normal, and a five second encounter can now land as 600 files. The fix is not to shoot less. It is to know what your fast mode actually records, to spend the buffer in short deliberate windows, and to shrink each sequence before you judge a single frame.

A five second burst is not five seconds of work

Hold the button for five seconds at 20 frames a second and you come home with 100 files. Hold it for the same five seconds at 120 and you come home with 600. The camera charges you nothing for the difference. Everything after the camera charges you for all of it: the offload, the second copy, the preview build, the scroll bar, and whatever attention you have left at eleven at night.

The distinction that keeps this manageable is between a capture opportunity and a deliverable opportunity. A bird does not perform 120 different pictures in a second because the sensor took 120 samples of it. A high rate raises the odds that one specific instant exists on the card. It does not create a hundred instants worth showing anyone.

So none of what follows is an argument for shooting less. It is an argument for spending the frame rate where the behavior earns it, and for arriving home with readable sequences instead of confetti. Which frame in a run finally wins is a separate craft, and it gets its own section near the end.

What the fast modes actually record, on three bodies we checked

The headline number on a spec sheet is often not the number you end up shooting. At the top of the range most bodies ask for something back: a smaller file, a cropped sensor area, JPEG instead of raw, or a burst the card cannot sustain. That matters at selection time, because an 11 megapixel JPEG is a timing document, not a print master. Here is what three manufacturers state on their own pages.

Specifications checked August 2026 on each manufacturer's own page. Firmware moves these numbers, so read yours before the shoot rather than after it.

If your camera is not in that table, that is not a judgment on it. It means we did not open its manual, and we would rather send you to check than print a figure we guessed. The habit worth copying is the checking. Know what your fast mode gives up before you use it on something that happens once a season.

The buffer is a time budget, not a frame count

Frames are the wrong unit in the field. Seconds are the right one. Canon quotes 93 raw frames at its 30 fps electronic rate, which is close to three seconds of holding the button before the camera starts to slow down. Sony's roughly 390 JPEG at 30 fps is closer to thirteen. Divide the burst depth your maker publishes by the rate you have set and you get the only number that matters while a subject is moving: how long you can actually hold.

Then shoot to that number on purpose. One to three seconds, release, look, go again if the behavior is still running. Short bursts protect the buffer, they keep you tracking the subject instead of watching a counter, and they land in the catalog as units you can read. A single fifteen second hold produces one enormous undifferentiated run with the interesting half second buried somewhere around position 400.

Silent capture makes the discipline harder than it used to be. There is no mechanical clatter telling your hands that 300 pictures just happened, so the feedback has to come from the frame counter and the buffer indicator instead of your ears. If you have never watched yours fill during a real sequence, that is a five minute test in the yard, not something to discover during a strike.

Choose the rate for the behavior, not out of anxiety

Canon publishes that whole ladder of electronic shutter rates for a reason, and most bodies have some version of it. Most photographers set one when the camera comes out of the box and never think about it again. The top of the ladder is not a default. It is a tool for a predicted event.

A practical split: run your ordinary continuous rate for the approach, the arrival and the follow through, and reach for the fast mode only when the thing you came for begins. A takeoff, a dive, a prey handover, a bat meeting a ball. A bird turning its head on a perch does not need 120 frames a second, and a portrait subject certainly does not. This is the same reasoning behind keeping a high bar for what you mark at the game, which we set out in in-camera rating for sports.

Pre-capture changes the calculation more than the top rate does. If the body is already holding the moment before your full press, Sony's one second, Canon's 15 shots, Nikon's Pre-Release Capture, then you no longer have to hold the button pre-emptively for ten seconds and hope. You react, and the frames before your reaction are already there. Check what the mode does to file type and buffer on your camera first, then save it for the moments it was built for.

Two more electronic shutter checks belong in the same test session. Fast pans and fast subjects can skew on bodies that read the sensor progressively, and some artificial lighting produces banding in silent modes. Neither is a reason to avoid the mode. Both are reasons to have seen it happen somewhere that does not matter.

Ingest so the pile arrives as sequences

A high frame rate day only becomes unmanageable when it lands as one undifferentiated folder. Two habits prevent most of that, and both happen before you look at a single picture.

First, two copies before any card is formatted, on two devices, verified by opening a few files from the second copy rather than trusting a progress bar. A fast shutter changes none of the arithmetic in our 3-2-1 backup workflow, except that you fill drives quicker and you are far more tempted to reuse a card in the field.

Second, keep chronological order and mark the boundaries. A run from 08:14:31 to 08:14:34 is one unit and deserves one decision, not 240 of them. Scatter those files into favorites folders during import and you lose the sequence context that tells you which frames belonged to the same pass. The general order of operations is in our ingest checklist, and for trips where the cards arrive faster than the evenings, the field version is in the wildlife trip workflow.

Then run the first pass on embedded previews wherever your software allows it. Building full raw previews for eight thousand files before you can begin is time spent rendering frames you were always going to pass over in a second. Save the deeper rendering for the small candidate set at the end.

Shrink the sequence before you judge any single frame

The mistake that eats an evening is going straight to 100 percent on frame one. At 120 fps, most of what is on the card is not a candidate, and inspecting it as though it were is how a five second encounter turns into an hour at the desk.

Work in this order. Cut the lead-in and the tail first, at fit to screen size: the framing setup, the focus acquisition, the empty sky after the bird left, the swing that finished two seconds ago. A 240 frame run routinely comes down to 60 on that pass alone, and none of it needed a judgment call.

Next, still at normal viewing size, scan for change. Head angle, wing state, the ball, the shape of the water, hands, the gap between the subject and the background. You are hunting for the ten or fifteen frames that differ in a way a viewer would notice, not the sixty that differ by a pixel of feather. Only then do you zoom, and only on those.

What happens next, which of the survivors wins, is its own craft and it is genre specific. The general ladder is in how to choose the best frame from a burst, and the wildlife version, eye first, then wing geometry, then behavior, is in how to cull a bird burst. There is no sense repeating either here.

Not every burst deserves the same attention

Split your sequences into two lanes before you start rather than after. One lane holds the runs with something at stake: rare behavior, the light you waited two mornings for, the play the client is paying for, the moment the whole trip was planned around. Those get a real comparison on a real screen, with a break before the final call. The other lane holds routine passes, distant flyovers, warm ups and obvious misses. Those get one sweep and a shortlist of one.

Keep a collection for the uncertain, high value runs and return to it later, when you are less attached to the fact that you were there. Fresh eyes are very good at noticing that ten files are one picture, and that the quiet frame you ranked third is the one carrying the moment.

Where software helps, and where it stops

Grouping is the part machines genuinely do well. Near duplicate detection clusters a run so it reaches you as one stack instead of 240 separate decisions, and a technical pass can flag the frames where focus clearly landed on a wing tip or a reed. That is real time off your evening, and it is first pass time, which is the cheap kind.

It is worth knowing what these tools were built on. Adobe's Assisted Culling, now in Lightroom Classic and cloud Lightroom, scores subject focus, eye focus and eyes open, and Adobe describes it as optimized for portraits and headshots. That is useful and honest of them to say: a 120 fps flight sequence in flat light is not the ground it was tuned on. We went through the wider gaps in what AI culling misses. For bursts the short version is that software is decent at technical rejection and grouping, and blind to which wing position tells the story.

Kepla is being built for the picking side of exactly this problem. It reads every frame from a shoot and hands back a shortlist, and it never deletes, moves or renames a single file, so the frame it passed over stays where you put it and is one tap away. The picking app for iPhone, iPad and Mac is still in development and we are not going to pretend otherwise. What is live today, and free while we build, is the Kepla booking page, which has nothing to do with frame rates and everything to do with the other half of the job.

FAQ