You’ve got one photo. Maybe it’s your gran at a birthday party in 1987, maybe it’s the wedding shot where everyone’s mid-laugh, maybe it’s your dog looking unreasonably pleased with himself in the garden. And you want it to move, just a little. Not a full film, just a few seconds where the person turns their head or blinks or the wind catches their hair. That’s it. That’s the whole ask.

Good news: this is genuinely doable now, with free tools, in about the time it takes to make a cup of tea. Bad news: there are a few ways to get it wrong, and one or two of them are legal problems rather than technical ones. So let’s go through it properly.

What you’ll be able to do

By the end of this you’ll have a short video clip, five to ten seconds, made from a single still photo, where the people in it move in a way that looks like them and not like a waxwork on a wobble. Faces stay recognisable. Nothing extra creeps into the frame. And you’ll know, before you even open a tool, whose photo you’re actually allowed to animate, because that bit matters more than which app you pick.

I’ll be honest, the animating part is the easy bit. The consent part is where people trip up.

Before you start

Right, the boring bit first, except it’s not boring, it’s actually the most important section in this whole guide.

You can animate your own photo. That one’s simple. If you took it, or you’re in it, or you own the rights to it, you’re fine. Where it gets trickier is other people’s faces.

If someone’s alive and they’ve said yes, fine, go ahead. If it’s a relative who’s died, and you want to animate an old photo of them for a family video, that’s not really your call to make alone; it’s a family decision. Somebody’s mum, somebody’s dad. Ask around first. I know that feels like a faff when you just want to make a nice clip for a memorial slideshow, but it’s fifteen minutes of conversation against months of someone feeling like their loved one was used without their say-so.

And never, ever animate someone else’s photo without their permission. Not a stranger’s wedding photo you found online, not an ex’s Instagram post, not a colleague’s LinkedIn headshot for a joke. Just don’t.

There’s also a proper legal point here, and it’s worth knowing even if you think it obviously doesn’t apply to you. In the UK, the Data (Use and Access) Act 2025 made it a criminal offence to create, or even ask someone else to create, a sexually explicit “purported intimate image” of an adult without their consent. That’s section 138 of the Act, and it adds a new section 66E to the Sexual Offences Act 2003. The government announced this on gov.uk on 7 January 2025. I’m not going to state penalties here, but the headline point stands: this is now a criminal matter, not just a “terms of service” violation. Worth knowing before you go anywhere near an AI tool with anyone’s photo but your own.

One more thing while we’re here: keep children’s photos out of third-party AI generators entirely. Don’t upload them to Kling, Runway, Pika, any of it. It’s just not worth the risk, and there’s no clip of your niece blowing out candles that’s worth uploading her face to a server you don’t control.

If the photo was taken by a photographer, professional or a friend with a good camera, technically the rights sit with them unless you’ve agreed otherwise. For old family snaps this is usually a non-issue, but for a wedding photographer’s shot, check first.

Get set up

Okay. Tools. Here’s what’s actually available right now, and what each one asks of you.

The simplest starting point, if you’ve got an Android phone, is Google Photos. There’s a “Photo to video” option in the Create tab that turns a still photo into a short clip, either with a preset effect or a text prompt describing what should happen. It’s free, though there’s a daily limit on how many you can generate, and a Google AI subscription plan raises that limit if you hit it. You’ll need a personal Google Account, you’ll need to be 13 or over, and it’s not available on Workspace or school accounts, nor in every region yet. There’s also an older effect called “Cinematic Photos” buried in there, which turns a flat picture into a moving 3D-ish video rather than animating people within it: different thing, worth knowing the difference.

Then there’s Gemini. Video generation inside the Gemini app now goes by “Gemini Omni”, which has replaced what used to be Veo access inside the app. It does image to video, can turn up to five photos into a single video, produces ten-second clips with native audio, and lets you keep editing across a conversation rather than starting from scratch each time. You’ll need a Google AI subscription (Google AI Plus or above), you need to be 18+, and what you get varies by tier and by country. Every video Gemini makes carries SynthID, an invisible watermark Google embeds, and Gemini itself can check a file for that watermark if you’re ever unsure where a clip came from. If you want the fuller studio version with more control, that’s Google Flow, at flow.google.com, running on credits tied to your plan.

Outside the Google ecosystem, Kling (kling.ai) treats image to video as a proper first-class feature, not an afterthought. I won’t tell you what the free allowance is or whether there’s a watermark on unpaid output, because that changes and you should just check kling.ai/pricing on the day you’re using it.

Runway (runway.com) has a free plan with 125 one-time credits to get you started, and their paid plans advertise “no watermarks” as a feature, which tells you plainly enough that free output probably carries one. Their Gen-4 model generates video from a single reference image and is decent at keeping characters looking consistent across the clip, which is exactly what you want here. I’m not going to quote you prices because they change and I’d rather send you to the site than give you stale numbers.

Pika (pika.art) gives you all its models on the free plan with no watermark, but there’s no ongoing monthly credit allowance, you buy credit packs as you go.

And then there’s Sora, from OpenAI. It exists, it does video generation, and you should check openai.com/sora directly for what your specific plan includes, because I couldn’t verify the details confidently enough to state them here.

Wherever you upload your photo, Kling, Runway, Pika, Google, it’s going to their servers to be processed. That’s just how these tools work. Before you pick one, have a think about where the finished clip is actually going: a private family WhatsApp group is a different call to posting it publicly on Instagram.

Try it yourself

Right, the fun part. There’s a pattern that makes these prompts work, and once you’ve got it you can reuse it for any photo. Say who moves and how. Say what stays still. Tell it the camera doesn’t move. Tell it to keep faces exactly as they are in the photo. Give it a length, five to ten seconds is the sweet spot. And do one movement per clip, don’t try to get grandad waving AND the dog running AND the curtains billowing all at once. One thing. Keep it simple.

Here’s a prompt for a photo with two people in it:

Animate this photo. The woman on the left turns her head slightly and smiles; the man stays still; a light breeze moves her hair. The camera does not move. Keep both faces exactly as in the photo. Five seconds, natural, no added people or objects.

Here’s one for when you want the safest, most conservative version, good for a first attempt on any photo:

Turn this photo into a short video with subtle movement only: the people blink and breathe, the background stays still. Do not change faces, hair or clothing.

Expect to run it three or four times before you get one you’re happy with. That’s genuinely normal, not a sign you’re doing something wrong.

Check the result

Once you’ve got a clip, don’t just glance at it on your phone and move on. Watch it at full size, not as a thumbnail, thumbnails hide all sorts of sins.

Pause on a few frames and put them side by side with the original photo. Look hard at the faces: eyes, teeth, jawline. AI video tools are getting good but faces are still where things drift first. Check hands and fingers too, that’s the other classic tell, an extra finger or a hand that bends the wrong way. Look at clothing and the background, make sure nothing’s melting or warping at the edges. And check nothing’s appeared in the frame that wasn’t in your original photo, an extra person, an object that wasn’t there.

If you want a second opinion that isn’t just your own eyes, this prompt is handy, feed it a frame from your video alongside the original photo:

Compare this frame from the video with the original photo. Has anything about the faces, hands, clothing or background changed? List every difference, however small.

If it doesn’t work

Face drifting or looking subtly wrong partway through the clip? Shorten it, ask for “subtle movement” specifically, and keep the camera still. Movement plus a moving camera is where most of the weirdness comes from.

If you’re getting proper warping, stretched features, background bending, try a different tool altogether, or crop the photo tighter around the people so there’s less background for the model to get confused by.

If the tool just refuses your photo outright, that’s usually a technical thing rather than a content issue: crop it to a standard ratio (square or 16:9 usually works), and strip out any text overlays or date stamps sitting on the image.

And if you keep getting a visible watermark you don’t want, either try the Google Photos route or accept that a paid plan on your chosen tool is the trade-off for a clean clip. Bear in mind Google’s output carries an invisible SynthID mark either way, which is a feature, not a flaw: it is how anyone can later check the clip was AI-made.

Questions people ask

Can I animate a photo for free?

Yes. Google Photos has a free "Photo to video" option in the Create tab, though there's a daily limit on generations. Pika gives you all its models on the free plan with no watermark, and Runway's free plan comes with 125 one-time credits.

Which AI tool is best for turning a picture into a video?

Google Photos is the simplest starting point on Android. Gemini does image to video with ten-second clips and native audio, Kling treats image to video as a first-class feature, Runway's Gen-4 keeps characters consistent, and Pika and Sora are also options.

Never animate someone else's photo without their permission. In the UK, the Data (Use and Access) Act 2025 made it a criminal offence to create, or ask someone to create, a sexually explicit "purported intimate image" of an adult without consent.

Can I animate an old photo of a relative who has died?

That's not really your call to make alone; it's a family decision. Ask around first, it's fifteen minutes of conversation against months of someone feeling their loved one was used without their say-so.

How do I check the animated video looks right?

Watch it at full size, not as a thumbnail, and pause on frames side by side with the original. Look hard at eyes, teeth and jawline, check hands and fingers, watch for melting clothing or background, and make sure nothing has appeared that wasn't in the original photo.

Keep exploring

If this has got you interested in AI video more broadly, have a look at “How to get AI to make a video”, which covers generating clips from scratch rather than animating an existing photo. If you want to go the other way and start with text or an idea rather than a photo, “How to get AI to create an image” is the place to start. And if your clip needs trimming, captions, or joining to other footage once it’s made, “How to get AI to edit a video” picks up from here.

Sources and review notes

https://support.google.com/photos/answer/16763021?hl=en-GB https://gemini.google/overview/video-generation/ https://gemini.google/subscriptions/ https://flow.google.com/ https://kling.ai/pricing https://runway.com/pricing https://runway.com/research/introducing-runway-gen-4 https://pika.art/pricing https://openai.com/sora/ https://www.legislation.gov.uk/ukpga/2025/18/section/138 https://www.gov.uk/government/news/government-crackdown-on-explicit-deepfakes

Review date: 2026-09-22