Lightroom Open Eyes Fixes Blinks With Your Own Photos, and That Changes How You Should Shoot Groups
Adobe's new blink fix can borrow open eyes from the frames you shot a second earlier. Used well, it saves the group photo. Used lazily, it invents a face. Here is how to tell the difference.
Every group photo has a blinker. Twelve people, one frame where the light is right and the toddler is looking at the camera, and Aunt Ruth caught mid-blink. For most of photography's history the fix was a head swap: find a frame where Ruth's eyes were open, cut her face out, line it up, mask it in, and hope nobody notices the slight change in angle.
Lightroom now does a version of that for you. Open Eyes, which Adobe shipped on November 10 as an Early Access feature in Lightroom on desktop, finds closed or half-closed eyes and generates open ones. That part sounds like every other AI retouch button. The interesting part is a dropdown most people will skip past.
It is called Reference, and it decides where the new eyes come from.
The dropdown that matters
Adobe's help page lists three choices:
- Nearby photos, the default: frames you shot around the same time.
- Choose photos: you pick the reference frames yourself.
- None: "Generate open eyes without reference photos."
Those three options are three different kinds of edit, even though the button looks the same.
With Nearby photos or Choose photos, Lightroom has real pictures of that person's open eyes to work from. Adobe's blog describes it as replacing closed eyes "using a reference image of that person from your library." Lightroom still generates the new eyes, but it generates them with real photos of Ruth to go on, which puts the result much closer in spirit to the old manual head swap. With None, Lightroom has nothing but the closed eyelids in front of it. It has to imagine what Ruth's eyes look like. Sometimes it will guess well. My expectation is that sometimes it will not, and that the eyes will not quite match hers. The people who know her best will notice first.
That is my position on this tool: Open Eyes is a reference tool that happens to have a generate button, not the other way around. The quality of the fix depends far more on what you shot than on anything you do in Lightroom.
Why the reference changes everything
Eyes carry more identity than almost any other feature in a portrait. The shape of the lids, how much white shows, the slight asymmetry everyone has, the way one eye squints a little more when someone smiles. A generic "open eye" pasted onto a familiar face reads as wrong even when nobody can say why.
Adobe's own advice points at this. The help page says to "use photos of the same person with their eyes open and with a similar face angle and lighting." Read that as a shooting instruction, not just an editing tip. If the only frame where Ruth's eyes are open has her head turned 30 degrees and lit from the other side, the reference cannot help much. If you have three frames from the same second, same pose, same light, it has everything it needs.
So the feature rewards an old habit: shooting in bursts. A professional at a wedding already does this for group shots, firing five or six frames in a row because somebody always blinks. Open Eyes turns those "extra" frames from backups into raw material. You keep the one frame where everything else is right, and the blink frames' neighbors donate the eyes.
How it finds the blinks in the first place
Adobe says Open Eyes "also powers part of Lightroom's Assisted Culling workflow," which scores each photo as likely open, likely closed, or can't tell. Lightroom 10.0 also added Faces panel scores that show "Eye sharpness and Eyes open scores and details for each face in a photo," per Adobe's what's new page.
That pairing is the real workflow. Culling finds the blinks across a whole shoot. Open Eyes repairs the keepers. And if you shot a tight burst, the "Nearby photos" default has good donors to choose from without you hunting.
What a good fix looks like
It helps to know what you are checking for before you start pressing buttons. A good blink fix is boring. The eyes point where the head points. The catchlights, those small bright reflections of your light, sit in the same spot as in everyone else's eyes. The lids have the same weight and crease as in the donor frame. The skin around the eyes keeps its texture instead of turning smooth and plastic.
A bad fix usually fails on one of those, and almost never on all of them at once, which is why it slips through on a quick look. Eyes that look slightly past the camera, a missing or doubled catchlight, lids that are too symmetrical: none of these jump out on a phone screen, and all of them show up in a print on someone's wall.
Put this into practice
Here is a routine for the next event or family session you shoot.
1. Shoot the group in bursts of five or more. Same framing, same light, no zoom changes between frames. You are building a reference set, so consistency matters more than variety.
2. Cull with the eye scores first. Use Assisted Culling to flag frames with closed eyes, and the Faces panel to see who blinked in each frame. Pick your hero frame on everything else: expressions, posture, the kid actually looking at the lens.
3. Open the hero frame in Detail view. Adobe's steps: select Remove, then the Eye Correction icon, then Open Eyes.
4. Choose the faces. Under "Apply to these faces," tick only the people who blinked. Leave everyone else alone. There is no reason to let a generator touch a face that was already fine.
5. Leave Reference on Nearby photos, then check what it picked. If the burst was tight, this is usually the right choice. If the nearby frames have a different head angle or the person was half out of frame, switch to Choose photos and pick the closest match by hand.
6. Select Generate, then zoom to 100 percent. Look at the eyes at full size, not on the grid. Check catchlights first: the little reflections of your light source should sit in the same place in both eyes and match the other faces in the frame. Mismatched catchlights are the fastest tell of a fake eye.
7. Compare against the donor frame. Put the fixed photo next to the reference you used. If the eyes look like the same person in both, you are done. If they look like a cousin, try a different reference or go back to a manual composite.
8. Use None only as a last resort. If you truly have no frame with that person's eyes open, None will generate something. Treat its output as a draft and show it to someone who knows the subject before you deliver it.
Where it falls down
It is Early Access, and Adobe has not published the fine print. The help page calls it "available for testing and feedback." Adobe's own blog post the same day calls it "public beta," so even the label is not settled. None of the four Adobe pages describing it (the help page, the what's-new page, the Assisted Culling page and the blog post) states whether Open Eyes uses generative credits, how many, or whether it needs an internet connection. Before you build a wedding delivery workflow around it, run it on a few dozen frames and watch your credit balance.
Adobe only describes it on desktop. The help page says it is available "in Lightroom on desktop" and does not mention mobile or web. If your editing happens on an iPad or phone, check before you plan around it.
Turned heads and changing light defeat the reference. The whole method depends on a donor frame that matches. A candid shot where the person was moving, or a frame lit by a flash that did not fire on the next shot, gives Lightroom a poor match. Expect odd results on profiles and strong three-quarter turns.
None is a different kind of edit. Without a reference, the tool is inventing eyes. That is fine for a background guest in a crowd shot. It is a bigger problem for the bride, the main subject of a portrait, or anyone whose face will be studied for years.
It is not acceptable everywhere. Photojournalism, documentary work, contest entries with no-compositing rules, and some clients' contracts all draw a line at altering what was in front of the camera. Even a reference-based fix moves pixels from one moment into another. Know your context before you press Generate, and disclose when the work calls for it.
Small faces may not be detected. Adobe's page describes the tool finding closed eyes automatically, but it does not say how small a face can be. In a large group shot from far back, check that every blinker actually appears under "Apply to these faces."
The skill is in the shooting
It is tempting to read Open Eyes as one more reason to care less about the moment of capture. I think it says the opposite. The tool works best when you gave it good material: a tight burst, steady light, the same angle across frames. It works worst when it has to make something up.
So the next time you line up a group, fire a few more frames than feels necessary. You are not wasting card space. You are shooting the reference set that lets the hero frame survive the one person who blinked. Then open Lightroom, pick that frame, and see whether the eyes it gives back still look like the people you photographed. You are the only one in the room who can judge that.