How to Turn a Product Photo Into a Video With AI
You can turn one product photo into a short video with an image-to-video model: the photo becomes the first frame and a one-line prompt says what moves. It is how every ppl.studio video is made, so below is what we have learned from rendering them: which photo to start from, the four kinds of product clip that work, the prompts that keep the product intact, and what each model costs per second.
2 min read
How image-to-video works
An image-to-video model gets two inputs: a still image and a text prompt. It keeps the still as the first frame and predicts the frames after it, following the prompt. Unlike text-to-video, which invents the whole scene, it starts from your real product, which is why it is the method for product content. The cost is that the model only knows the side of the product the photo shows, and it redraws every frame, so small text and fine detail are where clips fail.
Four product clips that work
- The orbit. The product frozen in one moment while the camera circles it once. Good for a hook and for a product page loop. See the 360° orbit clip workflow.
- Product in use. A person lifts, applies, pours or wears it. Start from a still of the person already holding the product, so the model doesn't have to put it in their hand. The product demo video workflow builds these.
- Talking about it. A creator-style person says a short script to camera with the product in shot. This needs a model that keeps speech: the creator UGC video workflow.
- The scene loop. The product still, the world moving around it: steam from a mug, light crossing a counter, rain on a window. Calm, cheap, and hard to get wrong.
Motion prompts that keep the product intact
- One motion, said plainly: “the camera slowly pushes in”, “she turns the jar to face the camera”.
- Tell the camera what to do, including “the camera holds”. Some models push in hard by default.
- Name only what the frame holds. “No logo” or “no text” tends to produce one.
- Keep the background colour by reference: “the same colour as in the first frame”, not a colour name.
- Small moves near the label. The bigger the motion across text, the more it warps.
The same rules apply to writing the still itself; see scene prompt.
Models and cost per second
| Model | Best for | Price per second |
|---|---|---|
| MiniMax H3 Max Turbo (fal) | Silent product shots, B-roll, many clips at once | $0.015 (480p), $0.04 (768p) |
| Gemini Omni 1.1 Flash | A person talking to camera, with the voice | ≈ $0.10 (720p) |
The numbers, speed and failure cases behind this table are in MiniMax H3 vs Gemini Omni Flash. For the stills themselves we use Nano Banana 2.1, which keeps the product and the person the same across shots.
Common failures and the fix
- The label warps: start from a sharper, closer still and keep the motion away from it.
- The product changes shape: the prompt asked it to turn further than the photo shows; limit the rotation or start from another angle.
- A person's mouth moves in a silent clip: faces looking into the lens start “talking”; have them look at the product.
- Made-up text appears: the prompt mentioned a message, a sign or a notification; describe a glow or remove it.
A still turned into a clip
Another: a pet moment
Sourceschecked November 1, 2026
Frequently asked questions
Can AI turn a photo into a video?
Yes. Image-to-video models take a still as the first frame and a short text prompt describing the motion, and generate a clip of a few seconds that starts from that exact picture. Gemini Omni, MiniMax H3, Kling, Veo and Runway all do it.
What is the best AI to turn a product photo into a video?
It depends on the clip. For a silent product shot or B-roll, MiniMax H3 Max Turbo is fast and cheap and stays close to the first frame in our renders. For a person talking about the product to camera, Gemini Omni Flash keeps the voice. Our measured comparison is in MiniMax H3 vs Gemini Omni Flash.
How much does it cost to make a video from a photo with AI?
Per second of video: about $0.015 at 480p and $0.04 at 768p on MiniMax H3 Max Turbo through fal, and about $0.10 at 720p on Gemini Omni 1.1 Flash. A 15-second product video built from three 5-second clips is roughly $0.60 to $1.50 in model cost.
Why does the label change in my AI product video?
The model redraws every frame, and small text is the hardest thing to keep. Start from a photo where the label is large and sharp, keep the motion small near the label, and avoid prompts that make the product rotate past the side the photo shows.
How long can an AI video from a photo be?
One generation is usually 4 to 10 seconds. Longer videos are several clips joined, each starting from its own still, which also lets you change the shot the way an editor would.
Turn your product photo into a video
Upload your product, pick a workflow, and ppl.studio makes the stills, animates them and cuts the video for TikTok, Reels and Shorts.
Browse the workflowsFounder of ppl.studio. Building AI tools for product marketing teams who need visual content at scale without the production overhead.