“Alice, animate my photo”: what it does and where it stops

The fastest way to see your photo in motion today is to ask Alice, Yandex’s voice assistant. The feature is free, takes about thirty seconds and needs nothing but an app you quite possibly already have. This chapter covers what exactly it does, where its abilities end and what to do next if one button is no longer enough.

What Alice does

According to Yandex’s help page, the feature works like this: you attach a JPEG or PNG of up to 10 MB, describe what should happen, and half a minute later you get an MP4 file. The resolution is 480p (854 × 480 pixels), the length is up to four seconds, and the file has no sound. The aspect ratio comes from the source photo. The feature is free with a limit on the number of generations; a paid “boost” that adds HD generations for a month is sold separately.

It is a very good product for what it is: a mass-market button, designed for someone to try once, show their friends and move on. It deliberately offers no settings, and that is exactly why it is so easy to use.

Where the ceiling is

The top three rows are about the scale of the job. The fourth, fifth and sixth are about different jobs altogether.

Is four seconds a lot or a little?

Honestly: for an animated photo, four seconds is usually enough. One movement with a start and an end fits in 4–6 seconds, and after that the model starts inventing ways to fill the frame, so a second movement nobody asked for appears. The difference between four seconds and fifteen is not the difference between “bad” and “good”, but between “an animated photo” and “a small scene with a story”.

No sound, on the other hand, is not a matter of size. A silent clip cannot be posted anywhere as it is: something has to be laid under it as a separate step. Sound created together with the picture, wind, footsteps, the hum of a room, matches the motion in the frame, because it was composed from the same description of the scene. A track laid on top afterwards cannot do that.

Speech is a separate route, not a setting

The most common disappointment sounds like this: “I animated a portrait, so how do I make it say a line now?” You cannot, because it is not the button next door but a different processing route. A portrait that moves goes into the model as the first frame. A portrait that speaks goes in as a reference, and the line is written in words in the request. The two cannot be combined in one request.

What the second route looks like in practice is covered in the chapter a talking photo.

When to move on, and when not to

Stay with Alice if the job is one photo, once, for yourself or to show in a chat. There is no reason to pay for that.

It makes sense to go further when one of these comes up:

• You need sound in the file: the clip is going to be published, and you do not want to add music separately.
• The person in the photo has to say something: a greeting, a character’s line, a short message.
• You have two photos and need a transition between them: before and after, an empty room and a furnished one, face and profile.
• You need a higher resolution, because the clip is going to a bigger screen, not a messenger.
• You need a series with the same character across several shots, and then to put it together into one file.

You can try any of these right here: attach a photo and describe what should come alive.

What to read next

How the whole method works and what the result really depends on: animate a photo with AI. Where the free tier ends at services like this: animate a photo free and online. The dose of motion that makes a clip look natural, or on the contrary “AI-made”: animating a photo.

In the Twelver chat a photo comes alive with sound, up to fifteen seconds long, and if you like with a spoken line in your own voice. Your first clip is earnable without paying: sign up, complete five first tasks and come back on the second and third day.