Most people shopping for "image to video" are holding a photo and a vague sense that it should move. That sense covers both jobs, and the two jobs cost completely different things — one is usually already included in something you pay for, the other is metered every time you press the button.
The same words, two different products
The confusion is not your fault. Both categories use the same three words on the box, and both produce an mp4 from a still image. What differs is whether new pixels get created.
The one-detail test: ten seconds, no account
You do not need to read a feature list to find out which job a tool does. You need one detail that must survive the clip.
Pick something in the photo that would be embarrassing to get wrong: a wedding ring, a logo on a shirt, a mole, the shape of a chair leg, a line of text on a label. Then watch that one thing at the first frame and again at the last.
That single check settles most of the confusion in this category. If the detail survives, you are looking at a motion tool and you do not need to buy anything new for job A. If the detail is re-drawn, you are looking at generation, and the price you are quoted is per attempt, not per photo.
What job A can actually do
Moving a still is not a lesser version of generation; it is a different craft, and it has real range:
- Camera moves. Push in, pull out, pan across, tilt. The photo stays the photo; the frame moves over it.
- Timing. How long each still holds, and where the cut lands relative to the beat.
- Layering. Two stills over each other, a slow dissolve, a wipe.
- Text and titles. Captions that arrive with the beat, which is most of what makes a photo sequence read as a video.
What it cannot do is add information. A closed eye stays closed. A blurry hand stays blurry. A photo taken from behind never shows the face. No amount of panning creates the other side of the room.
What job B adds, and what it charges for it
Generation adds the thing you could not photograph: the subject turns, the camera keeps moving past the edge of the frame, the scene continues. That is the entire value, and it is why the category exists.
The cost of that value is that every frame is a guess. The model has no memory of your ring beyond what it inferred from one image, so it re-invents the ring 24 times a second. When the guess is stable, the clip looks real. When it drifts, you get the classic tells: hands with too many fingers, text that becomes decoration, a face that ages three years in two seconds.
That is also why job B is billed per attempt rather than per photo. You are not buying the video, you are buying attempts at the video, and the attempt that finally works is the one you keep.
Why this decides your bill before you compare prices
If your job is A, the cheapest correct answer is usually a tool you already have. Every phone photo app can pan across a still, every video editor can keyframe a zoom, every design tool can put a photo on a timeline. You are not shopping, you are looking for a menu item.
If your job is B, no existing tool substitutes, and the comparison that matters is not price per clip but price per usable clip — the rate multiplied by how many attempts it took. A service that charges half as much and needs three attempts costs more than one that charges full price and lands it first time.
The hybrid most people should run
Once you can tell the two apart, the useful move is to stop choosing between them:
- Keep everything that must stay true in job A. Faces, products, text, logos, anything a viewer could check against reality. Move it, do not generate it.
- Spend generation only on the two or three seconds that need something new. The turn, the reveal, the entrance. Everything around it can be a still that pans.
- Join them at a cut, not at a morph. A hard cut from a moved still into a generated shot hides more error than a cross-dissolve ever will, because the viewer's eye resets at the cut.
This is also the cheapest structure, because the generated seconds are the only ones you pay per attempt for.
The checklist before you pay for anything
- Name the detail that must survive. If you cannot name one, you do not yet know what the clip is for.
- Run the ten-second test on the tool you already have. Most people stop here.
- If you still need generation, decide how many seconds genuinely need new frames. That number is your budget, not the clip length.
- Only then look at price, and look at it as price per usable clip.
Where this page stops
No tool names and no prices appear here on purpose. App menus move between versions, and a feature list written today can be wrong before this page ranks. The distinction between moving a photo and generating frames does not move, and the one-detail test works on any tool you open, this year or next.
Related on this site: what five seconds actually costs, choosing the frame before you upload, and the rest of the guides.