Three Different Things People Mean by "Animate Photos"

Someone tries a photo animator, gets a slow zoom across their picture, and concludes AI photo animation is overhyped. Someone else feeds a precious family portrait to a generative model and gets back a stranger wearing their grandmother's dress.

Both used a tool that said "animate photos." Neither got what they wanted.

The phrase covers three techniques that share almost nothing beyond the word. Most roundups list twenty tools without saying which technique each one uses, so you pick by star rating and find out afterward.

The three techniques

| | What moves | Good for | Fails at | | ---------------------------- | ------------------------------------------------------------------ | ------------------------------------------------- | ----------------------------------------------------------------- | | Pan and zoom (Ken Burns) | The camera. Nothing in the photo. | Landscapes, group shots, slideshow backgrounds | Faces. The subject stays frozen while the frame drifts. | | Face animation | Eyes, mouth, small head turns. Everything else holds. | Portraits where the person must stay recognisable | Anything below the neck. Full-body photos barely change. | | Generative video | Whatever the model decides: body, hair, background, clothing folds | Movement, dancing, expressive motion | Faithfulness. It invents, and invention drifts from the original. |

Pan and zoom is the oldest and the most oversold. It is a crop moving across a still image, which is why "animated" landscapes look fine and "animated" people look inert. Nothing about the subject changes.

Face animation is what MyHeritage's Deep Nostalgia popularised in 2021. It maps a small set of facial landmarks and drives them with a reference motion. The output stays close to the source because the model is deliberately not allowed to redraw much.

Generative video is the current wave, and it is a different kind of tool. It generates new frames rather than warping existing pixels. That is why it can make someone dance and why it can also make them not quite themselves.

Which one you want

The person must stay recognisable. Face animation, or generative video kept at its most subtle setting. This is the case for a photo of a relative, especially one who has died. Fidelity beats spectacle every time here.

You want motion that is obviously an effect. Generative video, turned up. Dancing pets, a selfie set to a beat, an avatar. Nobody is checking whether the face is exact, and the whole point is that it moves a lot.

Neither the subject nor the effect matters much. Pan and zoom. It is cheap, fast, and fine behind a title card.

The mismatch that produces most disappointment: wanting the first, using the third at full strength.

A trend worth knowing about

Search interest in animating old photos specifically has fallen sharply, down roughly 70% year over year, with "deep nostalgia" down more than 80%. Interest in animating photos generally has held steady.

The technique did not get worse. The novelty wore off, and the tools got absorbed into general photo-to-video products. If you tried this in 2021 and found it stiff and short, the generative models available now are a different experience: longer clips, more natural motion, and settings that let you dial back how much changes.

Doing it well

The settings matter less than the source photo.

Resolution on the face. More pixels there means more for the model to work with. A face cropped from a group shot at 100 pixels tall will look soft no matter which tool you use or what you set.

Front-facing beats profile. A model has to invent the far side of a face shown in profile, and invention is where results turn strange.

Nothing over the eyes. Sunglasses, deep shadow, a hat brim. These remove the landmarks every technique depends on most.

Less motion than feels right. Subtle reads as alive. Heavy reads as an effect. If a result feels uncanny, try the same photo at a lower setting before you try a different photo. The uncanny reaction is almost always too much movement, not the wrong tool.

When not to

Not every photograph wants to move.

Some pictures work because of the stillness: a considered portrait, a formal group photo, an image whose composition is the point. Adding motion can make them look like a screensaver.

And for a photo of someone who has died, watch the result alone before you show anyone. It lands differently than you expect, and the difference is not always welcome. If it feels wrong, that is a real response, not a technical failure. Leave it still.

Common questions

Is any of this free? Pan-and-zoom tools generally are, because the compute cost is near zero. Face animation and generative video cost real money per clip to run, so free tiers exist but are limited. Watch for watermarks. That is the usual trade.

Which technique does [tool] use? Look at their examples. If the face never changes expression, it is pan and zoom. If the face moves but the body does not, it is face animation. If the whole frame moves, it is generative.

Can I animate a photo on my phone? Browser-based tools work on phones without installing anything. The export is an MP4, which saves to your camera roll like any other video.

Does animating damage the original? No. Every tool works on a copy. Keep the original scan regardless. It is the artifact, and the video is one interpretation of it.

---

If you want the third kind, real generated motion from a single still, that is what our AI Live Photo Maker does, free to start and without a watermark. For a faded family portrait where staying recognisable matters more, Animate Old Photos is tuned for restraint.

Also worth reading: how to animate old family photos without making it strange.