AI Animation Tested on 5 Historical Photo Types (2026)
Cabinet cards, daguerreotypes, snapshot prints, group portraits, water-stained photos: I tested each historical photo type with AI animation and recorded what actually came back.
Family History Writer, Incarn
In short
Cabinet card portraits from the 1880s-1930s animate well with no preparation needed: good resolution, clear faces, three-quarter poses work naturally. Daguerreotypes and tintypes require careful flat scanning to avoid metallic glare, but the facial detail is often excellent once digitized properly. Snapshot prints from the 1900s-1950s vary widely: a sharp one is fine, a soft or tiny one benefits from AI upscaling first. Group portraits need individual face crops before uploading. Water-damaged or foxed prints animate cleanly if the face area is intact; otherwise a quick digital restoration pass fixes the worst of it. First animation free on Incarn (preview with watermark; EUR 1.99 for the HD version without watermark).
TL;DR: Cabinet cards and formal studio portraits from the 1880s-1930s animate well with no prep. Daguerreotypes need a flat scan to avoid glare. Snapshot prints vary: upscale the small ones first. Group portraits need individual face crops. Water-damaged prints work if the face area is intact, otherwise restore it first. Try your first animation free on Incarn, EUR 1.99 per photo to keep the HD version without watermark.
A genealogist I know animates every ancestor he finds in the archives. His nephews, who had never once opened the family tree, now browse it just to see the faces move. After watching him work through several boxes of family photos, I noticed the results were inconsistent in a specific way. Some photos came back looking natural on the first try. Others needed adjusting. The ones that needed adjusting almost always fell into recognizable categories.
Old family photo collections are not homogenous. A single shoebox might hold cabinet cards from 1895, snapshot prints from 1942, a tintype that probably predates your oldest living relative, and a water-stained group portrait no one has identified. Each format was made differently. Each digitizes differently. Each gives the AI animation model a different kind of input.
I ran tests on five historical photo types to see what actually comes back. Here is what I found.
Quick comparison
| Photo type | Typical challenge | Result | Action needed |
|---|---|---|---|
| Cabinet cards (1870s-1930s) | Formal three-quarter pose | Good to excellent | Usually none |
| Daguerreotypes and tintypes (1840s-1880s) | Metallic glare on scan | Excellent if digitized flat | Rescan if glare present |
| Snapshot prints (1900s-1950s) | Small size, possible blur | Variable | Upscale small or blurry copies |
| Group portraits | Faces too small in frame | Poor on full image | Crop to individual faces |
| Water-damaged or foxed prints | Staining, creasing | Good if face is intact | Restore face area if damaged |
1. Cabinet card portraits (1870s-1930s)
Cabinet cards are the most common format in European and North American family archives from the late 19th and early 20th century. A photographer mounted an albumen or gelatin silver print on a thick cardboard backing, usually around 4 by 6 inches. The resulting prints are often surprisingly sharp for their age.
The good news for anyone who wants to animate them: cabinet card photography was studio work. Photographers arranged their subjects with care, used good lighting, and worked at a large negative size. The result is typically clear facial detail, a recognizable expression, and a three-quarter pose that works well with AI animation models.
In my tests, well-preserved cabinet cards animated on the first upload in the majority of cases. No restoration pass, no upscaling, no special preparation. The main variable was condition. Cards with minimal yellowing or surface oxidation gave the best results. Cards that had been stored badly, with deep surface foxing or emulsion damage, needed more attention.
The formal pose that photographers used throughout this era, usually a slight turn of the head, natural chin angle, expression focused forward, is exactly the kind of input AI animation handles well. The model has enough facial structure to generate realistic movement.
One note on scanning: use a flatbed at 600 DPI or higher. Some families photograph cabinet cards with a phone rather than scanning them. This works but introduces the risk of ambient glare on the surface. Flat on a scanner avoids that entirely.
2. Daguerreotypes and tintypes (1840s-1880s)
This is where scanning technique matters most.
Daguerreotypes and tintypes both have a metallic base. A daguerreotype is a silver-coated copper plate that acts like a mirror: tilt it and you see a reflection, not a portrait. Tintypes (also called ferrotypes) are similar in structure, made on thin iron. Both are physically fragile and both create glare if photographed at the wrong angle.
The mistake I see most often: a family photographs a daguerreotype at a slight angle with their phone, chasing the angle where the image is most visible to the naked eye. The resulting digital file has metallic sheen across part of the face. The AI animation model has to work around incomplete facial information and the result is predictably uneven.
The fix is straightforward. Scan the plate face-down on a flatbed scanner, or photograph it flat on a neutral grey surface using diffuse light. No direct flash, no overhead angles. When properly digitized, daguerreotypes often have exceptional facial detail. The 1850s and 1860s studio photographers who made them were technically precise people working with slow exposures that required their subjects to hold still. The resulting images are frequently very sharp.
In my tests with well-scanned daguerreotypes, the animation results were among the best I recorded from any format. The facial detail gave the model a lot to work with. The historical stiffness of the poses, often criticized as unnatural, actually helped: the face is usually fully visible and angled toward the camera.
If you have a daguerreotype and a smartphone, do not photograph it like a print. Find a flatbed, or find someone who has one. It makes a significant difference.
3. Snapshot prints from the early-to-mid 20th century
Brownie cameras, early 35mm films, and the small-format snapshot photography that defined family archives from roughly 1900 to the 1960s produce a wide range of print quality. At one end: sharp, well-lit portraits at 2.5 by 3.5 inches that animate cleanly with no preparation. At the other end: soft, slightly blurry prints where the face is a small blur in a wide scene.
The album went from the attic to the cloud. The dust went from the cardboard box to the algorithm. And it scans right along with the portrait if you rush.
For this category, the determining factor is whether the face is large enough and sharp enough in the digital file. The face should be at least 200 pixels tall in the image. For a 2.5 by 3.5-inch print, this means scanning at 400-600 DPI. Phone photographs of these small prints work in good light if you hold the phone steady and fill the frame, but blur from handling is a common problem.
The cases where I found upscaling most useful: very small original prints (under 2 by 2 inches), prints that were slightly soft due to camera movement or poor focus, and prints from early 1900s roll film that had natural grain. AI upscalers, the kind used before animating rather than instead of animating, add visible facial detail. The guide to improving old photo quality with AI covers which tools handle this well for different input quality levels.
For sharp, well-exposed snapshot prints, none of this applies. Upload and test. The best AI tools for animating old photos gives an overview of tools and what each does better or worse for this input type.
4. Group portraits
Group photography introduces one problem specific to AI animation: the faces are small.
A formal family portrait from 1920, twelve people arranged in three rows in front of a house, gives the model a face the size of a postage stamp per person if you upload the full image. At that scale, there is not enough facial information for a convincing animation of any individual. The result is usually a collective swaying that looks more like a crowd simulation than a family coming to life.
The solution is to crop. Take the full group portrait, isolate one face per upload, and submit each separately. A five-person group generates five individual crops. Each animates well because each crop gives the model a face that fills most of the frame. The album photo is passed from the attic to the cloud, and each face gets its own moment.
This is not faster. Five uploads instead of one. But the difference in result quality is large enough to be worth it for any photograph that matters. For a group portrait you want to show the family, it is the approach that works.
The same logic applies to any photograph where the face is a small part of a large scene: a portrait taken from a distance, someone photographed at a table in a large room, a small figure in a garden. Crop to the face before uploading. The model animates what you give it; give it a face.
5. Water-damaged and foxed prints
Water damage is probably the most common form of physical deterioration in family photo archives. Prints stored in attics, basements, or garages encounter condensation, humidity cycles, and sometimes actual water contact. The damage shows as brown staining, tide marks, warped paper, or foxing (the brown spots that appear from mold spores).
The key question for animation is where the damage sits relative to the face.
Staining, foxing, and yellowing that affect the background, the clothing, or the margins of the print rarely affect the animation in any visible way. The model focuses on the face; what happens in the margins has little impact on the result. In my tests, prints with heavy foxing on the edges and clothing but clean face areas animated as well as undamaged prints of similar resolution.
Damage across the face is different. A crease through the eye, a water tide-mark obscuring part of the mouth, a stain over the cheek: these are cases where the model has to work around a gap in the information it needs. The animation often looks slightly off in that area.
The practical fix for face-area damage is targeted digital restoration before uploading. You do not need to restore the whole photograph, only the face. A quick pass in a restoration tool handles most common damage types in a few minutes. The result is often visually cleaner than the original print. What to restore and which tools handle specific damage types is covered in the old damaged photo guide.
For prints where the damage is on the face and you are unsure whether restoration is needed, the fastest test is to upload as-is first. The first animation on Incarn is free, which makes it practical to check the result before deciding whether to invest a restoration pass. If the result looks clean, no restoration was needed. If the problem area is visible in the animation, restore and re-upload.
A note on black and white tones
All five of the photo types above are black and white or near-black and white (sepia, brown tones, slight warm tones from albumen chemistry). This is not a problem for AI animation. Modern models work directly from grayscale and tonal contrast, not from color. The animated clip comes back in the same tones as the input: sepia in, sepia out. For a full look at how AI handles black and white photographs, the black and white photo animation guide covers the mechanics and options.
What the tests found, overall
Format matters less than condition and face size. A well-preserved cabinet card from 1890 animates better than a blurry snapshot from 1960. A properly scanned daguerreotype with excellent facial detail gives better results than a modern print of a small, soft image.
The two preparation steps that consistently made the most difference: a flat, glare-free scan for metallic or reflective originals, and individual face crops for group photographs. For everything else, the best approach is to test: upload the photo, check the result, and decide from there whether a restoration pass or upscale adds anything worth the extra step.
Your first animation is free at Incarn. No subscription. EUR 1.99 per photo to keep the HD version without watermark.
Frequently asked questions
Can you animate a daguerreotype photo with AI?
Yes, once it is properly digitized. Daguerreotypes have a mirror-like metallic surface that creates glare when photographed at an angle. Scan flat on a flatbed scanner at 600 DPI or higher, or photograph the plate on a neutral grey background with diffuse lighting. A good digital copy of a daguerreotype often has excellent facial detail and animates very well.
What old photo types animate best with AI?
Cabinet card portraits from the 1870s-1930s tend to give the best results: professional studio photography, good resolution, clear frontal or three-quarter poses. Early 20th-century snapshot prints work well when sharp and not too small. The common factor is a clear face taking up a reasonable portion of the frame.
Do group portraits animate well with AI?
Not as one image. When every face is a small element in a wide frame, the model has less facial detail to work with per person. Crop to a single face first, submit each person separately. The individual crops animate well; the full group shot usually does not.
Can I animate a water-damaged old photo?
Often yes. Water staining, foxing, and yellowing that affect the background or clothing rarely affect the animation. Damage across the face itself, particularly creases through the eyes or staining over the mouth, is more visible in the result. A targeted restoration pass on the face area before uploading usually solves it.

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.