What you can make with it
Two inputs, one output. Photo to video takes an image you upload and puts it in motion: the subject stays recognizable while the camera moves and hair, cloth, smoke and water shift. Text to video starts from nothing and builds the shot out of your description. Both hand back a 5 second MP4 at 768p with no logo burned into the corner.
Three formats are available: 16:9 for YouTube and landscape embeds, 9:16 for Reels, TikTok and Shorts, and 1:1 for feed posts. Choose before you generate. Cropping a finished 16:9 clip down to vertical throws away most of the frame and usually slices the top of a head off.
The model runs on a server and the result comes back to your tab. There is nothing to install, no extension, no plugin, no render job on your own machine. An old laptop and a new phone produce the same file because your hardware never touches the model.
| Setting | What you get |
|---|---|
| Length | 5 seconds |
| Resolution | 768p |
| Formats | 16:9, 9:16, 1:1 |
| File | MP4, H.264, plays anywhere |
| Watermark | None |
| Audio | None, the clip is silent |
| Typical wait | 30 to 90 seconds |
Silent output is normal for this kind of model. Add music or a voiceover afterwards in any editor, including the one on your phone.
Which photos animate well and which fail
The input photo decides most of the result. A model can only move what it can read, so a clean, legible image gives it room to work and a cluttered one gives it guesses.
| Works | Fails |
|---|---|
| One clear subject, centered or close to it | Collages and grids of several photos |
| Sharp focus on the face or the main object | Motion blur, heavy noise, low light phone shots |
| Even daylight or soft indoor light | Strong backlight that turns the subject into a silhouette |
| Faces at least a couple hundred pixels tall | Tiny faces in a wide crowd shot |
| Plain or simple background | Screenshots, memes, posters, anything full of text |
| Full head and shoulders inside the frame | Subjects cut off at the edge of the image |
Text is the most reliable way to ruin a clip. Letters have no stable shape for the model to hold on to, so a sign, a logo or a caption inside the photo will smear over five seconds. Crop it out before you upload. Old scanned family photos usually work better than people expect, as long as the scan is sharp and you crop tight on the person instead of feeding the whole album page.
Resolution matters less than sharpness. A 900 pixel wide photo that is properly in focus beats a 4000 pixel one shot through a shaky hand.
How to write the motion prompt
The prompt describes movement, not the picture. The model already sees the photo. Telling it that the woman has brown hair wastes the prompt. Telling it what the camera does and what moves is the whole job.
Three rules cover most of it. Use a camera verb (push in, pull back, pan left, tilt up, orbit, static shot). Ask for one action, not a sequence. Do not contradict yourself: a locked off camera and a slow dolly in the same sentence produce a jitter, because the model tries to satisfy both.
| What you want | What to write |
|---|---|
| A slow zoom toward the face | Slow push-in on the face, camera moves forward, background stays still |
| Life in the photo without moving the person | Static camera, hair and clothing move slightly in a light breeze |
| A reveal of the landscape | Camera pulls back and slightly up, revealing the valley behind the subject |
| Water that actually moves | Waves roll in slowly, foam spreading on the sand, locked off camera |
| A cinematic look | Slow dolly left, shallow depth of field, warm late afternoon light |
| The subject to react | Subject turns head slightly left and smiles once, camera still |
| A product to look expensive | Slow orbit around the bottle, soft reflections moving across the glass |
Short beats long. Two clauses is usually enough, and past about twenty five words the extra detail starts fighting itself. If a clip comes out wrong, change one thing and run it again instead of rewriting the whole prompt. Our step by step walkthrough of a single photo is in the guide on how to animate a photo with AI.
What free means here
A set number of clips per day, currently two, with no sign up and no card. No email, no confirmation link, no trial that quietly turns into a charge. The counter resets the next day.
The output is a real file. No watermark, no logo, no bar across the bottom, no 480p downgrade for free users. You download the MP4 and use it wherever you want, including commercially. Paid credits for people who need more than the daily allowance will come later on this site.
What is not included: audio, clips longer than 5 seconds, 1080p or 4K, batch jobs, and an API. Those are real limits, not upsell bait. If your project needs a two minute video with dialogue, this is the wrong tool and no prompt will fix that.
Realistic limits of AI video today
Every generator on the market shares the same failure modes. Anyone showing you only flawless demos is showing you the winners out of many attempts.
| Limit | What to do about it |
|---|---|
| Clips are about 5 seconds | Plan a cut, not a scene. Generate two or three clips and join them in an editor. |
| Hands and fingers warp | Crop above the hands, or ask for a still body and a moving camera. |
| Text and logos turn to mush | Keep them out of the input photo. Add the text back on top afterwards. |
| No lip sync | Do not prompt speech. Use a closed mouth or a slight smile. |
| Physics goes wrong | Ask for small movements. Slow beats dramatic in a 5 second clip. |
| Faces drift over time | Shorter, smaller motion holds a likeness better than a big camera move. |
The practical approach is to treat every run as cheap and disposable. The first attempt is a test of the photo, the second is a test of the prompt, and the third is usually the one you keep.
What people use it for
Social posts are the obvious one: a still product shot becomes a scrolling stopper in 9:16 without hiring anyone. Product mockups come next, where a slow orbit around an object reads as a studio shot even when the source is a photo taken on a kitchen table.
Animating old family photos is the use that surprises people. A grandparent's portrait with a gentle push-in and a small head turn gets a reaction that a static scan never gets. Storyboards for pitches are another: five seconds per beat is enough to sell a sequence to a client without paying for animation.
Thumbnails and covers also count, since you can pull a single frame out of the MP4 and use it as a still. The image to video tool is the same generator with the photo mode already selected if that is all you need.
Browser tool, desktop app, or subscription
Three ways to make an AI clip, with different costs before you have seen a single result.
| Browser tool | Desktop app | Subscription service | |
|---|---|---|---|
| Cost to try | Free daily clips | Download, install, sometimes a licence | Card details are often required first |
| Account | None for the free clips | Usually an account plus the install | Yes, with email confirmation |
| Watermark | None | Common on free tiers | Common on free tiers, gone when you pay |
| Wait time | 30 to 90 seconds | Depends on your GPU, minutes upward | Fast on paid plans, queued on free ones |
| Resolution | 768p | Limited by your hardware | Higher tiers unlock higher resolution |
| Works on a phone | Yes | No | Usually, through an app |
A desktop app makes sense when you generate hundreds of clips and already own a strong GPU. A subscription makes sense when you need long videos, audio and higher resolution every week. For everything else, the browser wins on the only metric that matters at the start: seeing a result before you commit anything.
Privacy
Your photo is uploaded, sent to the model, used to make the clip, and then deleted. It is not kept in a gallery, not used for training here, and not shown to anyone else. The finished MP4 is available to you for a short window so you can download it, then it goes too.
Nothing is posted anywhere on your behalf. There is no public feed of what people generate. Because there is no account for the free tier, there is no profile to link your clips to in the first place.
One rule worth stating: do not upload photos of other people who have not agreed to it, and do not upload images you have no right to use. Upload a photo, write one line about the motion, and see what comes back.