Pet product demo clip on Seedance 2.5: treat, toy and leash prompts

For pet products, keep the animal doing one simple thing and the product clearly visible. Seedance 2.5 on Sume prices a 10-second 9:16 clip at $5.78 at 720p, and a 15-second clip at $8.67.

Animals are harder for a video model than a still object. Fur, paws and chewing motion can look wrong. Plan on more than one take, and do the first ones at 480p.

Why pets need more takes

A dog chewing or a cat batting a toy involves fast, small motions with contact between a body and an object. Those are the moments a generated clip is most likely to blur or deform. Shorter clips give fewer such moments, so a 10-second demo with one clear action is better than a 20-second tour of everything the toy does.

Three prompts

Each uses a generic animal and the product as a reference photo.

• Treat: "A golden retriever sits on a kitchen floor and looks up. A hand offers one round treat from the reference photo; the dog takes it gently and wags its tail. Handheld phone framing at dog height."
• Toy: "A tabby cat on a living room rug bats a feather wand toy from the reference photo twice, then pounces. Soft daylight, camera low and steady."
• Leash: "A person in a gray coat walks a small terrier on a retractable leash from the reference photo along a park path. Camera follows from behind at knee height, steady pace."

Price by length

Pet clips are short.

Request for the treat clip

The treat photo goes in input_references; the reference images for this model are capped at 9 by Sume's validation.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: pet-treat-001" \
  -d '{"model":"seedance-2.5","prompt":"A golden retriever sits on a kitchen floor and looks up. A hand offers one round treat from the reference photo; the dog takes it gently and wags its tail. Handheld phone framing at dog height.","input_references":[{"type":"image_url","image_url":{"url":"https://example.com/treat.jpg"}}],"duration":10,"resolution":"480p","aspect_ratio":"9:16"}'

Setting up the shot

Get the animal's attention cheaply in the prompt: say where it looks and what it does first. A dog that "looks up at the hand" gives the model a clear gaze target, and a gaze target keeps the head steady. Shoot at the animal's eye level, since a camera from above makes pets look small and makes paws and tails harder to place. Keep the background plain, a kitchen floor or a rug, so the animal and product are the only moving things.

Claims and care

A generated pet is not a real pet. Do not use one to claim a food is eaten or a toy survives chewing, because the animal in the clip never did either. That matters for pet products, where buyers act on safety. Use the generated clip for mood and the scene, and use real footage of a real animal for any proof.

Keep the product's size honest. Models can enlarge a toy or shrink a leash handle to suit a composition. Compare the product to the animal in the output; if the toy looks twice its real size, reword the prompt with a size cue like palm-sized.

If the first draft has a distorted paw or an odd chewing motion, shorten the action to one beat, lower the camera motion, and retry. A 480p retry of a 10-second clip costs $2.69.

Pick a model tier

Use Mini or Fast for a first screen of the idea if you want to spend less, and move to 2.5 only for the keeper. The Mini price for the same 10-second 9:16 clip at 480p is $0.88. Animal motion on a cheaper tier is not a safe guide to what 2.5 will do, so treat the draft as a framing check only.