Prepare Your Old Photo for AI Animation: 5 Rules (2026)
Resolution, cropping, damage, face angle: five things to check before uploading an old photo to an AI animator. They decide whether the result looks real or robotic.
Family History Writer, Incarn
In short
The most important thing: the face needs enough pixels. A cropped, front-facing portrait with the face filling at least a third of the frame animates far better than a wide group shot scanned at 72 DPI. Damage across the eyes or mouth shows in the result. Sepia, faded, and black-and-white all work without colorizing first. Incarn, which we make, lets you test your first animation free (preview with watermark; HD without watermark once paid) at EUR 1.99 per photo.
TL;DR: The face needs pixels. A cropped, front-facing portrait with the face filling at least a third of the frame will animate better than a wide group shot. Damage across the eyes or mouth shows up in the result. Sepia and black-and-white both work without colorizing. Try your first animation free on Incarn, EUR 1.99 per photo to keep the HD version without watermark.
An amateur genealogist animates every ancestor he tracks down in the archives. His nephews, who had never once clicked through the family tree, now browse it just to watch the faces move. After a few dozen portraits, he started noticing a pattern. Some animations came back fluid on the first try. Others looked slightly off, something strange in the eyes, a movement that felt mechanical. The difference, almost every time, came down to the photo itself, not the tool.
There are five things worth checking before you upload. None of them require software expertise.
1. Resolution: the face needs pixels to work with
AI animation works from the pixels available in your photo. The model predicts what the face would look like a fraction of a second later, frame by frame. More facial detail at the start means more precise, more natural movement in the result.
A practical threshold: the face should be at least 200 pixels tall in the image. Below that, the model has less to work with and the animation looks softer, less convincing at full size.
For a physical print you're digitizing, scan at 600 DPI for a standard wallet-size or 4x6 portrait. That gives a digital file around 1400 to 2500 pixels on the short side, well above the minimum. For a larger framed print, 300 DPI is enough. The photo album moved from the attic to the cloud. The dust moved from the cardboard box to the algorithm, and it scans right along with the portrait if you rush.
Phone photos of prints work better than most people expect, as long as you fill the screen with the photo and hold the phone steady. Modern phone sensors have high enough resolution. The problem is almost always motion blur or the phone catching glare from the surface, not megapixels.
A small 1931 print, barely larger than a playing card, can animate well if it was scanned carefully. The size of the original print does not dictate the quality of the result. What matters is the resolution of your scan.
If your digital file is already small (under 800 pixels wide), a pass through an AI enhancement tool can roughly double the effective resolution before you animate. These tools reconstruct facial detail using AI and make a real difference on the hardest photos. The guide to improving old photo quality with AI covers the main options and when each one earns its place in the workflow.
2. Face size in the frame: crop before you upload
The face doesn't need to fill the entire image, but it should take up a meaningful portion of it. A useful rule: the face should occupy at least a third of the image height.
If you're working with a group shot, a wedding party, a family standing in front of a house, crop to a single face before uploading. Each face cropped separately will animate with more detail than the whole scene submitted at once.
You don't need to be aggressive about it. A portrait that shows the shoulders and a bit of background is natural and animates well. The goal is to avoid the situation where the face is a small element in a large frame and the model has to do most of its work on sky, walls, and clothing.
One practical tip for groups: if you want animations of multiple people from the same photo, create one cropped version per face and submit them separately. You get better results on each person than you would from a single upload of the full group.
Cropping takes two minutes in any photo app and makes a larger difference than most people expect.
3. Physical damage: what the model handles and what it doesn't
Not all damage is equal.
Stains, yellowing, fading, and foxing on the background, clothing, or edges rarely affect the animation in any visible way. The model focuses on the face; what's happening in the margins matters far less.
Damage across the face is a different story. A crease running through the eyes, a scratch crossing the nose, a water stain obscuring part of the mouth: these are the cases where the animation has to work around a hole in the information it needs, and the result often shows it as a slightly unnatural movement.
The practical fix: targeted restoration before you upload. You don't need to restore the whole photograph, only the face. A quick pass in Remini (mobile), jpgHD, or any AI restoration tool can clean up visible damage in the face area in a few minutes. Photos that look difficult as paper prints often animate cleanly once the damage is removed digitally.
The workflow for restoring old photos with AI covers which tools work best for which types of damage, including heavy creasing, chemical fading, and water marks.
For photos with deep damage across the face, the sequence that works best is: restore first, then animate. Trying to animate a heavily damaged portrait and hoping for a clean result is usually the slower route.
4. Face angle: why front-facing portraits give better results
The model generates movement in three dimensions from a two-dimensional image. It predicts what the face looks like turning, blinking, and shifting expression based on the information in the starting photo.
A face turned directly toward the camera gives the model the most to work with. Front-facing portraits animate the best.
Three-quarter portraits, the classic angle in most professional studio photography from the 1900s through the mid-twentieth century, work very well too. The slight turn often adds to the natural quality of the movement rather than reducing it.
True profiles, where only one side of the face is visible, are harder. The model has less structure to extrapolate from, and the animation is usually less convincing on the hidden side of the face.
If you have multiple photos of the same person and want the best animation, prioritize front-facing or three-quarter over profiles. This is the one factor you cannot fix after scanning. Worth thinking about when you're choosing which photo to use.
5. Tone: sepia, faded, and black and white all animate as-is
This surprises people, but it is consistent across tools: modern AI animation models work directly from whatever tonal profile the photo has. Sepia, grayscale, partially faded color, even the flat yellowish tone of very old albumen prints: all animate without preparation.
The animated clip comes back in the same tones as the input. Sepia in, sepia out. Black and white in, black and white out. The model does not add color or change the look of the photograph.
Colorizing first is an option, not a requirement. If you want the final animation in color, colorize the photo in a separate tool first, then animate. If the original tones are part of the point, animate as-is. Both give good results.
One situation where adjusting the tone is worth considering: very heavily faded photos where there is almost no contrast left. The model reads facial structure partly from tonal variation. If the photo is so flat that you can barely make out the face, a simple contrast boost in any photo editor (or a full colorization pass) before animating can help.
For more on what the model actually does with grayscale portraits, animating black and white photos with AI covers the mechanics in more detail.
What if the photo still needs work?
The two situations where preparation before animating consistently pays off:
Resolution is too low. If the face is smaller than 200 pixels tall, an AI enhancement pass is worth the few minutes it adds. The improvement in the final animation is usually clear.
Visible damage across the face. Creases, scratches, or stains on the eyes or mouth will appear in the animation. Cleaning up the face area digitally solves the problem at the source.
For everything else: upload and test. The first animation on Incarn is free, which makes it practical to try a photo before deciding whether it needs further work. Most photos with decent resolution and a clear face animate on the first attempt.
Quick checklist
Five questions before you upload:
- Resolution: is the face at least 200 pixels tall? If not, enhance the resolution first.
- Cropping: is the face taking up at least a third of the frame? If not, crop.
- Damage: is there visible damage across the eyes or mouth? If yes, restore that area.
- Angle: is the face roughly front-facing or three-quarter? If it's a profile, consider whether you have a better photo.
- Tone: is the image in color, sepia, or black-and-white? No action needed: all three animate fine.
Tick all five and you're ready. If you are unsure how a photo will animate, test it. The best AI tools to animate old photos gives an overview of the tools available and how each handles different types of old photographs.
Your first animation is free at Incarn. No subscription. EUR 1.99 per photo to download the HD version without watermark.
Family History Writer, Incarn
Rose writes Incarn's English-language guides on old family photos, archives and the AI tools that bring them back to life. She started with a shoebox of her grandparents' prints.