Why Your Photos Look Soft: Motion Blur vs. Missed Focus vs. Diffraction
Soft photos almost always trace to one of three causes: motion blur, missed focus, or…
Moving the focus point beats focus-and-recompose almost every time at a wide aperture. Recomposing after locking focus tilts the plane of sharpness off your subject by the angle you tilt the camera — invisible on the back screen, glaring at 100%. At f/1.4 that shift can eat the entire depth of field.
I learned this the expensive way on my Fujifilm XF 56mm f/1.2: a whole set of headshots that looked sharp on the screen and turned out front-focused on the cheekbone once I pulled them into Lightroom, because I’d locked focus on the eye and swung to recompose. That afternoon converted me. Here’s exactly why recomposing costs you sharpness, when it’s genuinely fine, and how modern autofocus has changed the calculus. It’s the focus lever in my sharp-photos system.

Focus-and-recompose is the old workflow born from cameras that had only a single, central autofocus point. You place the center point on your subject’s eye, half-press to lock focus, then — while holding the lock — swing the camera to frame the shot the way you want, and fire. It’s fast, it’s intuitive, and for decades it was the only practical way to get an off-center subject sharp with a center-point-only AF system. Millions of great photos were made this way.
The problem is geometry. When you focus, the camera sets the distance to a flat plane parallel to the sensor. When you then tilt the camera to recompose, that plane of sharpness tilts with it — and your subject, which sat on the plane a moment ago, now sits slightly in front of or behind it. The technique made sense when it was the only option. Today, when nearly every mirrorless camera lets you put the focus point anywhere in the frame, it’s often solving a problem you no longer have.
The culprit is what’s often called cosine error. Picture focusing on a subject dead center at, say, two meters. The camera locks the focal plane at exactly two meters, perpendicular to the lens axis. Now you rotate the camera to move the subject to the side of the frame. The subject is still two meters away in a straight line — but the flat focal plane has swung, so at the subject’s new position in the frame, the plane now sits at a slightly different distance. The subject falls off the plane by an amount that grows with how far you recompose and how close you are.
Whether that error matters depends entirely on depth of field. At f/8 on a wide lens, depth of field is deep enough to swallow the error and nobody notices. At f/1.2 on my 56mm, depth of field at portrait distance can be a couple of centimeters — and the recompose shift easily exceeds that, dropping the eye out of focus while the cheek or the ear stays sharp. This is exactly the “sharp in the wrong place” signature I describe in why your photos look soft: the frame isn’t blurry, it’s just focused a few centimeters off the mark. The wider your aperture, the more brutal the penalty.

It’s perfectly fine — sometimes preferable — whenever depth of field is deep enough to absorb the small shift. Shoot a landscape at f/8, a street scene at f/5.6 with the subject several meters away, or a group at f/8, and the recompose error disappears into the depth of field. In those situations focus-and-recompose is faster than hunting the focus point across the frame, and speed can be the difference between catching a fleeting moment and missing it.
The rule of thumb I use: the error scales up as the aperture opens and the subject gets closer. So focus-and-recompose is safe for distant subjects, stopped-down apertures, and wide focal lengths, and dangerous for close subjects, wide apertures, and long focal lengths — exactly the portrait conditions where it fails. Photographylife’s discussion of the technique lines up with what I see on my own files, and their fundamentals at Photography Life are a good cross-check. Know which regime you’re in, and you’ll know instantly whether recomposing is safe.
| Situation | Better method | Why |
|---|---|---|
| Portrait at f/1.4–f/2 | Move the focus point | Depth of field too thin for recompose error |
| Landscape at f/8–f/11 | Either works | Deep depth of field absorbs the shift |
| Street candid, distant subject | Focus-and-recompose | Fast; error is negligible at distance |
| Close macro subject | Move the point (or manual) | Razor-thin depth of field, any shift ruins it |
| Moving subject, eyes visible | Eye-detect AF | Camera tracks and holds the eye for you |
The objection to moving the focus point used to be speed, and modern controls have mostly killed it. A rear joystick — the little nub next to your thumb on most enthusiast bodies — lets you flick the point around the frame without taking the camera from your eye; it’s the fastest method once it’s muscle memory. Touchscreen focus-point selection is even quicker for many, letting you tap the eye directly, and on some cameras you can drag the point on the screen with your thumb while still looking through the viewfinder.
Pick a smaller AF area for precision — single point or a small zone — so the camera focuses on the eye rather than grabbing the nearest high-contrast edge. I keep single-point AF on the joystick for deliberate work and a wide tracking area for action. The muscle memory takes a week; after that, placing the point on an eye is faster than the old lock-and-swing ever was, and it’s dead accurate. The same precise-placement discipline carries over to manual focus in hard cases, which I cover in manual focus techniques for the night sky.
Largely, yes — for anything with eyes. Eye-detect autofocus finds the subject’s eye, locks onto it, and tracks it as they and you move, placing the focus exactly where the recompose method risks missing. On both my bodies it has quietly ended the front-focused-portrait problem: I frame however I like, the camera holds the near eye, and I fire. It’s the single most useful autofocus advance of the mirrorless era, and it’s why I no longer recompose for portraits at all.
It isn’t magic everywhere. Eye-detect can hunt in very low light, lose the eye when the subject turns away, or grab the wrong person in a crowd — and it does nothing for a landscape, a product, or any subject without a face. In those cases you’re back to placing the point yourself. But for the classic focus-and-recompose use case — an off-center person at a wide aperture — eye-detect is simply the better tool, and it’s worth learning your camera’s face and eye AF settings cold. For subject-separation work where this matters most, my background blur guide and portrait lens recommendations go deeper on shooting wide open safely.
Back-button focus deserves a mention here because it changes the whole ergonomics of focusing. By moving autofocus off the shutter button onto a dedicated rear button (AF-On, or the AE-L/AF-L button on my X-T5), you separate focusing from firing entirely. Press the back button to focus, release to lock, then shoot as many frames as you like without the camera refocusing — and refocus only when you choose. It ended a whole category of “the camera refocused as I pressed the shutter” misses for me.
It also makes a disciplined focus-and-recompose safer when you do use it, because you control precisely when focus locks and it never grabs again mid-recompose. Combined with eye-detect on the back button, it’s the setup I run every day: back button to acquire and hold the eye, joystick to override when I want a specific point, shutter purely for timing. Set it up once — the menu path is in every modern camera’s autofocus section — suffer the one-day learning curve, and you’ll wonder how you shot without it. Focus accuracy stops being luck and becomes a decision you make on purpose.
Turn on eye-detect AF and move your focus point with the joystick or touchscreen for anything at a wide aperture — stop recomposing for portraits entirely. Keep focus-and-recompose in your pocket for distant, stopped-down subjects where it’s fast and harmless. And set up back-button focus this week; it’s the ergonomic change that makes every focus decision cleaner. Sharp eyes at f/1.2 aren’t luck or a better camera. They’re a focus method you choose deliberately, one frame at a time.
Only at wide apertures and close distances. Recomposing after locking focus tilts the plane of sharpness off the subject by the angle you rotate the camera. At f/8 the deep depth of field hides this, but at f/1.4 the shift can exceed the entire depth of field and drop the eye out of focus.
Cosine error is the focus shift that happens when you focus on a centered subject and then rotate the camera to recompose. The flat focal plane rotates with the camera, so the subject ends up slightly in front of or behind the plane. The error grows with how far you recompose, how close the subject is, and how wide the aperture.
Move the focus point for portraits and any close subject at a wide aperture, where the recompose error is large relative to depth of field. Focus-and-recompose is fine for distant subjects at stopped-down apertures, where deep depth of field absorbs the small shift and the method is faster.
For subjects with visible eyes, largely yes. Eye-detect finds and tracks the eye and places focus exactly where recomposing risks missing, which has effectively ended front-focused portraits on modern mirrorless cameras. It does not help with landscapes, products, or any faceless subject, where you still place the point yourself.
Leave a Reply