What you actually get back: the length, resolution and sound of an AI video
The FAQ answers this four sentences at a time, so here it is in full. About five seconds, 720p, the shape of the photo you uploaded, and sound the model generated itself rather than a silent file. What that is genuinely good for, and what it is not.

The site’s FAQs answer this in four sentences at a time, scattered across a dozen pages. Here it is once, in full. A clip made from a photo in this app runs about five seconds, at 720p, in the shape of the photo you gave it, and it arrives with sound the model made up to match the scene. None of that is adjustable, and all of it is worth understanding before you generate your first one.
The numbers that do not move
Three things are fixed, the same for every action in the library, waving or breakdancing, a birthday clip or a save the date.
Five seconds. Every motion action returns a clip of about that length. There is no slider for ten or thirty, because the model behind this app is pinned to one configuration rather than offered as a set of options. What it costs to make an AI video covers the credit side of that same fixed shape, if you are weighing a clip against a restyled still.
720p. Not the sharpest number in consumer video, and the right one for where these clips actually get watched. More on that below.
The shape of your source photo. The clip is not generated into a fixed frame that you then have to crop into. It takes the aspect ratio of whatever you uploaded. A portrait photo returns a portrait clip, a square returns a square. This is the one fact that genuinely changes what you should do before you generate, not after: decide the shape you want the clip in by cropping the photo first, because there is no second chance to reframe the output.
The part that surprises people: it is not silent
Most tools in this category hand you a silent video and leave the sound to you. This one does not. The model invents audio to match the scene it is generating, footsteps on gravel, fabric moving, a murmur of a crowd, the particular quiet of an empty room. Waving comes back with the faint ambient noise of wherever the model decided you were standing. It is not a soundtrack. It is closer to the room tone a film crew records on set, present and mostly unnoticed, there to stop the clip feeling like a vacuum.
What it is not is music. If you are picturing a clip you can drop a song under, you are starting from ambient sound rather than silence, and the two behave differently once you try to layer something on top. A quiet track under a clip with its own ambient noise is usually fine. A loud one fights it. Worth a listen with headphones before you assume what you are working with.
What 720p is actually good for
720p is not a compromise so much as a match for where these clips live. A phone screen, a message thread, a post in a feed, a story that loops for a few seconds and moves on. At that size, on that screen, 720p looks completely clean, and a viewer has no way of knowing the file is not sharper than it is.
Put the same file on a television across a lounge room, or project it on a wall at a party, and the gap shows. The file is being asked to cover far more physical space than the resolution was built for, and it will look soft in a way it never does on a handset. That is a planning problem rather than a quality problem: know where a clip is going to be watched before you decide it is good enough, not after.
Why a longer clip is not simply the better version
It is a reasonable instinct to want more, ten seconds instead of five, a loop you could leave running on a screen all evening. The five second limit is not a corner cut to save cost. It is close to the edge of what the model can hold together.
Every frame in a generated clip has to agree with the one before it, and the face in frame ten has to still be the same person as the face in frame one. Why your animated photo looks wrong goes through what happens when that grip slips over a single five second clip, drift in a hand, a face that softens at the edges of a big movement. Stretch that same job out over thirty seconds and the model is holding a much longer thread of invention together, with more frames for a small error to compound across. Breakdance already asks a lot of a five second window, a full-body movement, consistent clothing, a floor that has to stay a floor. A longer version of the same clip is not obviously a better one, it is a harder one, and the honest five seconds is closer to where this kind of generation is reliable than a longer promise would be.
Working with what you actually get
None of this is a reason to avoid the format, it is a reason to plan around its actual shape rather than the one you assumed. Crop the photo to the aspect ratio you want before you generate, because that decision is made for you the moment you press the button. Watch the first result with sound on, because the ambient audio is part of what you are sending along, not an afterthought you add later. Keep the viewing distance in mind, a phone screen or a message thread rather than a wall, and 720p will never be the part of the clip that disappoints anyone.
Choosing a photo that will animate well is the fuller version of the planning problem, the crop, the light, the crowd size, all decided before a single credit gets spent. And if the occasion is a card rather than a single clip, holiday cards that move covers sending a five second file to people who were expecting a photograph, which is really the same question as this page asked from the other end: know what you are sending before you send it.
Common questions
- How long is a clip made from a photo?
- About five seconds, the same length for every motion action in the library. It is not a setting you choose, it is a property of the model this app is built on.
- What resolution do I get back?
- 720p. Fine for a phone screen, a message thread, or a social post. Not what you would choose for a television or a projector, where the file has to stretch much further than it was made for.
- Is the clip silent?
- No. The model generates its own sound to match the scene it invented, footsteps, fabric, a crowd, a room. That is not a music track, and it is worth knowing before you plan to lay music over what you assumed was silence.
- What shape is the video, portrait or landscape?
- Whatever shape the photo was. The clip is generated at your photo's aspect ratio rather than a fixed frame you have to crop into afterwards.
- Can I get a longer or higher resolution clip?
- Not from this app. Every action runs the same configuration, and that is a deliberate ceiling rather than a missing setting, for reasons the model itself explains once you understand how the five seconds gets made.

