Make an All-in-Focus Product Reference From Several Photographs
Make an All-in-Focus Product Reference From Several Photographs
You have a small product arrangement and one obstacle: the near face carries lettering you need to read, the far edge carries a detail you need to check, and no photograph you have holds both sharp. Not the shot from last week, not the one the client sent. Focus falls away before it reaches the second thing.
The fix is to stop hunting for one photograph. Make several, each holding a different slice of the depth in focus, then combine the sharp parts into a single reference. The result is a document that shows what the object looks like — near and far, in one view.
What it isn't is a photograph. The composite was built from four or five exposures at fixed positions under fixed light. It tells you what is printed on the tin. It cannot tell you that a single exposure, or a camera moving past the product, will ever hold that same depth of field. Those are different questions, and a stack answers only the first.
Decide which near and far details must be readable
"Everything sharp" is not a target. It's a wish, and it will not tell you where to put the focus or when you are finished.
Name the things instead. Two or three is usually right:
- The line of small text on the front, from the first character to the last.
- The stamped seam at the far rim.
- The corner where the wrap meets the base.
That list is your coverage map. Every named region has to be in focus in at least one source photograph. Anything you didn't name is allowed to be soft; if it turns out to matter, you name it and add a frame for it.
Then check that focus is actually your problem. Stacking recovers depth. It does not recover sharpness. If the only frame you have of the far edge is blurred from camera shake, or too low in resolution, or so dark it's mostly noise, a stack will reproduce all of that faithfully, one defect at a time. That situation calls for new source material, not a stacking guide.
One decision to make before you shoot: the composition is going to be identical in every frame, so freeze it deliberately. Same viewpoint, same camera height, same subject position, same lights. Say it out loud before you start, because the temptation during a long series is to nudge something and tell yourself it doesn't matter.
Capture overlapping focused regions
Mount the camera. Lock the lights. Set the object and leave it alone.
Then shoot frames whose sharp regions overlap. The overlap is the whole trick. If frame two's sharp band begins exactly where frame one's ends, you have drawn a line and hoped; the odds are good that the line misses and leaves a soft stripe nothing later can repair. Give every band shared ground with its neighbours.
Change the focus distance between frames rather than the camera position. Move the camera and you change the perspective — the far edge shifts against the near face, alignment has more to correct, and the composite begins to show it.
Expect small changes anyway. Focusing a real lens alters the image slightly even when the camera hasn't moved: the field of view can breathe, and the effective aperture shifts a little as you rack focus. That's why alignment exists as a separate step, and why it can't be perfect.
Watch for what moves while you aren't looking. A cable settles. A label lifts at one corner and drops back. A softbox drifts a centimetre on its stand, and a reflection that sat in one place is now in another. Alignment fixes position. It does not fix a reflection that moved because the light moved, or a shadow that lengthened between frames. If the highlight on the rim changes shape from frame one to frame four, you will get a smear where the two versions meet — and you will get it again in every correction you attempt from those frames.
More frames are not automatically better. Fifty shots of the same plane add nothing and slow everything downstream. A frame earns its place if it holds a region you named, or if it closes a gap between two frames that didn't overlap. Otherwise it's noise with a filename.
Keep the originals. All of them, untouched. They are the evidence for every decision you make later, and the only thing you can check a suspect mask against.
Align and blend, then inspect the masks
Load the photographs as layers in one document. Adobe's help page for the operation — Create a composite with extended depth of field, whose main procedure was last updated in February 2026 — sets out the route: align the layers, then apply Auto-Blend with Stack Images. The same page is explicit that Stack Images is not the same as Panorama blending. They are separate options, and only one of them is for differently focused regions. Choose the wrong one and you will be stitching a panorama that doesn't exist.
That is the documented procedure, and it is where most tutorials stop. It is not where you stop.
Alignment registers features from frame to frame. It does not put any depth into focus, and it cannot tell you whether your frames covered the object or skipped a band. A perfectly aligned stack with an uncovered stripe is still a stack with an uncovered stripe.
The blend then proposes a composite, choosing the sharpest available source for each region and building masks to switch between them. Treat that as a proposal, not a verdict. Open the result at 100 percent and look at the places you named at the start:
- The near lettering, character by character, including both ends.
- Every boundary where one frame's territory meets another's.
- The rear detail.
- Any glossy, transparent, or reflective edge, where selection tends to choose wrongly.
A soft seam disappears at thumbnail size. That is the difficulty. A doubled edge and a mask that landed in the wrong place both look like a clean image at any reduced view — a quarter size or smaller is usually where it hides — because your eye is averaging across the error. Inspect at 100 percent or don't inspect at all.
The working standard to hold yourself to is traceability: every sharp patch in the composite should be traceable to a named frame. If you can't say which exposure a region came from, you haven't verified it yet — you've just looked at it.
A stack with two soft patches
The rest of this section is a constructed example — a walkthrough of the method, not a session run for this article.
Say you're shooting a plain matte tin, no branding, about ten centimetres tall, sitting on a fixed surface. The camera is on a tripod, angled about thirty degrees off-axis and slightly elevated, so you see the near curve, the receding body, and the top rim. The near-to-far depth across the view is roughly seventy millimetres. Two soft lights, unchanged for the whole series. Near detail: a short line of small printed text low on the near curve. Far detail: a small stamped seam at the far side of the rim.
Four frames:
- F1 focused on the front lettering.
- F2 focused a short step behind F1. The lettering reads, though its lower edge has begun to soften; the depth out to about a third of the way back is sharp.
- F3 focused on the far rim.
- F4 focused slightly behind F3, to be sure the rim's back edge is covered.
You can already see the hole. Between the back of F2's coverage and the front of F3's, the middle of the tin's body belongs to no frame. F1 and F2 were aimed at the thing you cared about. F3 and F4 were aimed at the other thing you cared about. Nothing was aimed at the space between them.
You align and stack anyway, and the composite comes back. Two patches look wrong, and both look wrong the same way — a soft area with no texture in it. They are not the same problem.
First patch: the foot of the last character. The automatic mask drew its boundary along the base of the lettering and handed that strip to F2, where the base was already softening. Open F1 alone and the same strip is crisp, character and all. A source frame holds this detail. The mask chose badly, and that is repairable.
Second patch: the middle of the body. Open all four frames one at a time and look at that band. F2's rear edge is the best of them, and F2's rear edge is soft. F3's front edge is soft. Nothing in the series holds this region sharp.
This second one is worse than it looks, because the blend made it look fine. Forced to choose something, the composite picked the least-soft option across the gap and faded the transition smoothly, so there is no visible seam at the boundary — just an area that turns to mush when you zoom in. A thumbnail will not flag it. The only way to find it is to ask of this patch, and of every other one, a single question: which source frame has this sharp?
Repair from a real source, or recapture
That question sorts every soft patch into one of two piles, and the piles get different treatment.
A source frame holds it. Correct the mask so the composite uses that frame across the whole region. This is normal, and it isn't a sign you did something wrong. A photographer's own write-up of the process — Lincoln Harrison on focus stacking — describes exactly this habit: gathering focus coverage, running the blend, and then going back to fix the masks the automatic pass got wrong rather than accepting its first output. In the tin example, you would open F1, compare it against the composite at the foot of the last character, and paint the layer mask so F1 owns the character from tip to base. Then inspect again at 100 percent, the same way, because moving one boundary can expose another.
No source frame holds it. Here you have two honest options and no third.
The first is to recapture. Return to the same setup, restore the same lighting and the same object position, and make the missing frame — focused on the band nothing covered. Then rebuild the stack with the new frame included. If you kept the originals, the new frame has to match them, not the other way around.
The second is to ship the reference with the limitation stated. A reference that says "the middle of the body is not resolved in this version" is a useful document. A reference that quietly fills that band with texture borrowed from a neighbouring region is a false one. It would put detail into a document whose entire value is that it shows what was actually observed, and there would be no way for the person reading it to tell the difference.
Which is to say: don't synthesize the gap. Don't clone the band, don't content-fill it, don't paint in plausible grain. Every one of those operations produces something that looks like product information and isn't.
Two housekeeping steps to finish. Save the layered file, because the layer stack is the record of how the composite was built and the flattened image is not; anyone reviewing your work later, including you, needs the stack and the source frames together. And label what you deliver: a composite made from several exposures, not a single photograph. Put it in the filename, the caption, or the delivery note — wherever the person receiving it will actually read it.
What the composite does not establish
The finished file shows the near lettering and the far seam sharp in one view, which is what you needed, and it does that honestly, because every sharp pixel in it came from an exposure where that region was genuinely in focus.
Notice what the method depends on. It works precisely because no single frame covered the depth. Four frames, four focus positions, four sharp slices. The composite exists as a substitute for a photograph that couldn't be made.
So it can't be used as evidence about that photograph. If someone looks at the reference and asks whether the proposed moving shot will hold the same depth of field, the stack has no answer. A camera moving past the product brings a changing viewpoint, a changing focus distance, and usually different settings — a wider aperture, a shorter focal length, a closer working distance — and each of those changes the depth of field in ways that have nothing to do with how many frames you stacked. An extended-focus reference describes the object. It says nothing about what a lens will capture in one pass while it moves.
If that question needs answering, it needs a different test: a shot made under the proposed conditions, in the proposed motion, examined at the size the final image will be seen. That is separate work, and the stack is not a rehearsal for it.
What you are left holding is two things. A reference where the front lettering and the far rim can both be read, built from a known set of frames. And those frames themselves, still on disk, which are the only reason anyone can check the composite later. Keep them together. The composite can tell a client what is printed on the tin. Only the frames can tell you how you know.
Frequently asked questions
What should determine which details need to be in focus in an all-in-focus product reference?
Name two or three regions rather than wishing everything sharp, such as the line of small text on the front, the stamped seam at the far rim, and the corner where the wrap meets the base. Every named region must be sharp in at least one source photograph. Anything unnamed may be soft. Also check that focus is the problem; stacking recovers depth, not sharpness lost to shake, low resolution, or noise.
How should source frames be captured for focus stacking?
Mount the camera, lock the lights, set the object, and leave it alone. Change the focus distance between frames rather than the camera position, because moving the camera changes perspective. Give every sharp band overlap with its neighbors. Watch for moved cables, lifted labels, drifting lights, or changed reflections that alignment cannot fix. More frames are not automatically better; keep all originals untouched.
What does Auto-Blend with Stack Images do, and what still needs inspection?
The documented route is to align layers, then apply Auto-Blend with Stack Images, which is not the same as Panorama blending. The blend proposes a composite by choosing the sharpest available source for each region and building masks. That is a proposal, not a verdict. Alignment registers features but does not add depth or check coverage. Inspect at 100 percent, especially the named regions, boundaries between frames, rear detail, and glossy or reflective edges.
In the tin example, why are the two soft patches different problems?
The first patch, at the foot of the last character, can be repaired because F1 holds that strip crisp; the automatic mask chose F2 where the base was already softening. The second patch, in the middle of the body, is worse because no source frame holds it sharp; the blend smoothed across the gap, so it looks fine at thumbnail size but turns to mush when zoomed. That gap calls for recapturing the missing frame or stating the limitation, not synthesizing texture.
What does an all-in-focus composite not establish?
It shows what the object looks like with near and far details sharp in one view, but it is not a photograph and does not show that a single exposure or a camera moving past the product will hold the same depth of field. A stack describes the object. A question about a proposed moving shot needs a separate test made under the proposed conditions, motion, and viewing size.