Image to video: turn a product photo into a video ad
AI image-to-video, done inside Claude: ask in chat and Polaris turns a product photo into a UGC video ad. Model choice, prompts, captions.
Image to video means feeding a single product photo into an AI model that generates motion around it — turning a static shot into a few seconds of scroll-stopping video ad. The fastest way to do it with Polaris is right inside Claude: connect Polaris once through its MCP connector, then ask Claude in plain chat to turn your product photo into a UGC-style video ad. A live Polaris panel appears in the conversation and renders the finished clip by itself, routing your shot to a strong-fit video model (Google Veo, Kling, Seedance, Nano Banana, and more). Generate winners directly inside Claude — or in the Polaris web app if you prefer a prompt-first composer.
What "image to video" actually means
A product photo is one frozen frame. Image-to-video takes that frame and predicts what the next few seconds of motion would look like — the camera pushing in, the product rotating, a hand reaching for it, light shifting across the label. The result is a short clip that feels shot, not rendered.
For ecommerce, that is the whole game. Static images stop being enough the moment you are competing in a feed full of motion. A talking-head UGC clip, a slow product hero spin, or a quick "unboxing" beat all convert differently than a still — and now you can generate them from the exact photo already on your product page.
Two things separate a usable ad from an AI curiosity: the model you pick for the shot, and the prompt you give it. Polaris is built to handle both for you.
How image-to-video works in Polaris
The flow is deliberately short. You are not learning a video editor — you describe an ad to Claude and let the models do the heavy lifting.
- Connect Polaris to Claude once. Add the MCP connector (polarisugc.com/api/mcp) in Claude's connectors and authorize it with OAuth. You only do this a single time.
- Ask Claude for the ad. In plain chat, say what you want — "turn this product photo into a UGC video ad" — and attach your product image or paste a script.
- We pick the model per shot. Different models are good at different things, so Polaris routes each shot to a strong fit rather than locking you into one engine. You set model, aspect ratio, length, and sound by tapping option cards, not configuring a dashboard.
- Winners render in the chat. A live Polaris panel fills in progress and then the finished video by itself, usually in 20 to 180 seconds. Batches render in parallel, each with its own panel, and you can download them all as one ZIP.
- Caption and ship. Add one-click Auto-Captions in a TikTok-native style, then download and post.
Because it runs on credits, you can iterate — try a few directions on the same product and let the winners earn the spend. Failed renders auto-refund, so a miss never costs you, and you can see the full walkthrough in generating ads inside Claude.
Choosing the right model per shot
There is no single "best" video model, and pretending there is one is how you get flat ads. The honest answer is that each model has a personality, and the right choice depends on the shot. Polaris chooses per shot so you do not have to memorize the trade-offs — but it helps to understand them.
| Model | Where it tends to shine (in our experience) | Reach for it when |
|---|---|---|
| Google Veo | Coherent motion and prompt-following on more complex scenes | You want a believable person, environment, or camera move |
| Kling | Smooth, dynamic motion and physical realism | The product is moving, rotating, or being handled |
| Seedance | Fast, punchy short-form energy | You need quick, high-tempo UGC-style beats |
| Nano Banana | Precise image generation and product-shot control | You are crafting or editing the source frame before it moves |
The point of a multi-model studio is that a single ad rarely wants a single model. A hero spin, a lifestyle shot, and a talking-head testimonial each have a different best fit — Polaris matches the shot to the model instead of forcing everything through one.
Writing prompts that get usable ads
Image-to-video rewards specific, visual direction. Vague prompts produce vague clips. A few rules we live by:
- Name the camera move. "Slow push-in," "orbit around the product," "handheld follow." Motion direction is the single biggest lever on how the clip feels.
- Set the scene, not just the subject. "On a sunlit kitchen counter, morning light, soft shadows" beats "nice background." The environment sells the product as real.
- Say who is in frame. If you want UGC, describe the person and the action — "a woman holds the bottle to camera and smiles" — rather than leaving it to chance.
- Protect the product. Keep the label, color, and shape faithful to the source image. Describe the product accurately so the model does not drift.
- One idea per shot. Do not cram a spin, a pour, and a testimonial into one generation. Break the ad into shots and prompt each cleanly.
- Match the platform. Vertical framing, fast energy, and a clear first second for TikTok and Reels; a calmer hero look for a product page.
If you are not sure how to phrase it, describe the ad you have in your head to Claude in plain language — Polaris is built to take a rough idea and turn it into a real clip, then let you refine.
From one clip to a whole ad
A raw generated clip is the start, not the finish. Two features do most of the last-mile work inside Polaris:
- One-click Auto-Captions. We transcribe the audio word-accurate and burn captions directly into the video in 4+ TikTok-native styles — TikTok, Hormozi, Beast, and Neon. Burned-in captions are what make short-form ads watchable on mute.
- Recreate any winning ad. Paste a reference ad you admire and get your own version built around your product. It is the fastest way to turn a proven format into a first draft instead of staring at a blank prompt.
You can also generate fresh product-shot images inside Polaris, then animate those — so the image you feed into image-to-video is already on-brand.
Why teams pick Polaris
Polaris is an AI ad studio built only for ecommerce, and every feature points at one outcome: more ads that actually convert, faster.
- Multiple top models, chosen per shot. You are not betting your whole ad on one engine — we route each shot to a strong fit across Veo, Kling, Seedance, Nano Banana, and more.
- Captions that ship. Word-accurate, burned-in, TikTok-native styles in one click.
- Recreate proven formats. Turn a reference ad into your own version instead of guessing.
- It works inside Claude. Through our MCP connector, you generate ads directly in chat — the finished clip renders in a live panel, with nothing new to learn and no dashboard to babysit.
- Ecommerce-first and credit-based. Pro is $49/mo for 2,000 credits (roughly 66 video ads), and it gets cheaper on longer terms (15% off on 3 months, 25% off on 6 months). You iterate until the ad is right.
The reason to start from a product photo is simple: you already have one, and it is the most honest possible input for an ad. Image-to-video turns that asset you already own into motion that sells.
Get started
Ask Claude to turn your product photo into a video ad and watch the finished clip render right in the chat — or use the Polaris web app. Create your first ad or see pricing to find the plan that fits your volume.