Why this site exists

Image to video tools all sell the same promise and none of them tell you the one thing that decides the outcome: which photograph you feed them. A picture cropped at the chest cannot dance. A face in heavy shadow cannot act. Pick the movement the photograph already suggests and the result improves before the model does any work at all.

So this site documents every movement in AI Video, across every category, one page each, with a straight answer about the kind of picture each one needs. You should know up front that AI Video is ours. That is exactly why the limits are written down: a site that only listed the good parts would be an advert, and adverts do not earn trust.

Where the library comes from

The app reads its action catalogue from a public JSON file rather than shipping it inside the binary, so new actions can appear without an App Store release. This site reads the same file when it builds. One source, two consumers, no second copy to keep in sync. That is also why the clip length quoted everywhere is 5 seconds: it is the duration the catalogue's default model is configured for, not a round number we liked.

What we hold ourselves to

• The library is read, not writtenEvery action name, description and grouping on this site is pulled from the app's own catalogue when the site is built. We do not retype it, so it cannot drift out of date, and we cannot quietly add an action the app does not have.
• Every preview ships in the productThe still on each action page is the same image the app puts on that action's card. Nothing here is a render made to sell a feature that does not exist.
• No numbers we cannot countThe number of actions on this site is whatever the app's catalogue lists, read at build time and never typed into the copy. You will not find a success rate, a user count or a star rating here, because we do not have honest versions of those.
• The limits get their own sectionGenerated motion invents what it cannot know, hands are the weak point, and faces drift on big movements. That is written on the library page rather than buried.
• Photo guidance is craft, not a promiseThe advice on each page is about framing and light: what a picture needs to contain before an action is even plausible. It is not a claim about how the model will behave on your photograph.

Who publishes this

Photo Motion Lab is published by MomentumLab AI Pty Ltd, a small Australian software studio that builds and runs its own apps. There is no outside investor and no client work behind these pages, which is the reason we can afford to publish the section about what the tools cannot do.

Getting in touch

Corrections, arguments, and requests for an action we have described badly: [email protected]. If a page here is wrong we would rather hear it than leave it up.