Generate video from a product photo in Claude Code: a five-step recipe

1. Start from a photo you like

   Image-to-video keeps what is in the frame, so the photo decides the product, the backdrop, and the light. Ours was the studio shot from the GPT Image 2 guide: one sneaker, beige backdrop, no lettering. A photo with text or a busy background gives the model more to get wrong.

2. Describe the motion, not the picture

   The photo already says what the shoe looks like; the sentence should say what moves. "Slowly rotates while the camera pushes in" is a motion. "A white sneaker on a beige backdrop" is a caption, and the model will invent motion to fill it.

3. Draft on the cheapest tier

   Seedance 2.0 Mini at 480p costs $0.07 per second (480p) — a five-second draft is small change. Iterate the prompt there. Spend on Seedance 2.0 or Kling 3 only for the version you ship, and only once the cheap tier is doing what you asked.

4. Check the result against the photo

   Play the clip next to the original and look for what changed. Our first draft drew a two-letter logo on the tongue that was not in the photo; adding "unbranded, no logos, no text, no lettering anywhere" to the prompt removed it. The second draft kept the shoe clean but turned the gum sole white; adding "keep the gum-coloured rubber sole exactly as in the photo" fixed that. The third run is the one below.

5. Ship it, or send it up a tier

   A 480p draft is fine for a social post or a product-page thumbnail. For a hero video, run the same prompt through Seedance 2.0 at 1080p or 4K, or Kling 3 for a second look — the agent reuses the photo and the sentence, so the only thing that changes is the price on the quote.

The run that worked, end to end

That is a real Claude Code session, unedited, and the clip above is what it produced: the shoe from the photo, the gum sole intact, no invented lettering, a slow turn and a push-in. Three drafts on the cheap tier, three small charges, and the prompt that finally worked is the one you can copy from the block above.

How the session decides what to run

Claude Code writes and runs code; rendering pixels is not something it does. The run went through Actionway's video generation capability, and the Skill installed with the CLI is what taught the agent to price the job, run it, wait, and bring the file back. Before replying, it read the capability index the Skill keeps on disk, which is why the first message already had the plan: which capability, three models, what each costs per second. The options are presented neutrally — no default, no recommendation; say "you pick" and it takes the best-quality tier. Choosing one is the approval, so the call runs at the shown price without a second question, and if the wallet cannot cover it the agent stops before anything is charged and hands you a top-up link with the amount prefilled.

Video is priced per second of output, and the resolution tier moves the price. The Seedance family covers the range — Seedance 2.0 Mini from $0.07 per second (480p) for drafts, Seedance 2.0 from $0.14 per second (480p) up to 4K — and Kling 3 from $0.20 per second (720p) gives the same brief a second look. The same photo also became three ad versions through the image service, which is the e-commerce remix use case end to end. There is no subscription; one balance covers every capability in the catalog.

Set it up once

npm install --global @actionway/cli@latest
actionway init
mkdir -p "$HOME/.claude/skills"
cp -R "$HOME/.actionway/skill/actionway" "$HOME/.claude/skills/actionway"

You need Node.js 20 or newer and an Actionway account; after sign-in the agent tells you your wallet balance and, if it is empty, links you straight to top-up. Prefer a plugin or a Remote MCP server to the command line? Both end at the same catalog and the same wallet — the get started page lists the steps per client side by side.