Input: product attributes, platform, and desired action
Output: a reusable prompt and a matched model setting
Aspect: 9:16, 1:1, or 16:9
Build log
1. Name the protected fields
Product shape, label text, logo position, color, and material. For example: keep the front label unchanged and do not redraw packaging text.
2. Choose one variable
Hand pickup, water droplets, table scene, kitchen counter, or shadow move. Two variables usually cause the first failure.
3. Set a camera and duration
Use one of these starts: still product, slow 15-degree turn, or locked tabletop camera. Then pick 5-8 seconds for the first test.
4. Add the correction rule
Say what to do when the model invents a package or label: keep the exact reference, remove extra props, or return to the master photo.
5. Test at the lowest cost
A 5-second Agnes Video test is 10 credits. Upgrade only the winner to a paid render: 90 credits for 5 seconds of Seedance 480p or 195 credits for 720p.
Complete execution record
Task boundary and source gate
Treat AI product video prompts as a deliverable, not a definition. The job receives product attributes, platform, and desired action and owes a reusable prompt and a matched model setting in 9:16, 1:1, or 16:9. Write the rejection reason first, then turn it into the first hard constraint.
Do not start until the source photo has a visible product code, a note on material and color, the approved claim, and a named target module. Check packaging damage, sticker state, lighting direction, and background rights before generation. Do not invent certifications, competitor claims, price windows, or a model using the product in an unsafe way. If the seller cannot prove it, the image does not carry it.
The imgmov route
Open Asset Library first, attach the master SKU as the locked reference, then build the module or shot around that reference. Keep offer copy, platform text, and CTA outside the visual prompt. In imgmov, the approved image becomes the source of truth and generation supplies only the requested scene, angle, or motion.
Keep platform requirements outside the model prompt. The prompt can describe a scene; a named checklist records the crop, claim, consent, and console check that make it deliverable.
Lock the variables that cannot drift
Lock product silhouette, label spelling, logo position, material, primary color, accessory count, contact shadow, and the platform safe area. Scene, props, and motion may change; these fields may not.
Put the locked fields at the top of the prompt, not at the end: Use a fixed sentence order: protected product, locked camera, one variable, duration, and failure correction. Start with Agnes Video at 2 credits/second; move to Seedance 2.5 at 18 credits/s for 480p or 39 credits/s for 720p only after the concept passes. Save that prompt with the reference, ratio, and model so the next run starts from a decision instead of a guess.
Step-by-step build path
1) Name the protected fields: Product shape, label text, logo position, color, and material. For example: keep the front label unchanged and do not redraw packaging text. Leave one checkable artifact from this step; do not start the next until it exists. 2) Choose one variable: Hand pickup, water droplets, table scene, kitchen counter, or shadow move. Two variables usually cause the first failure. Leave one checkable artifact from this step; do not start the next until it exists. 3) Set a camera and duration: Use one of these starts: still product, slow 15-degree turn, or locked tabletop camera. Then pick 5-8 seconds for the first test. Leave one checkable artifact from this step; do not start the next until it exists. 4) Add the correction rule: Say what to do when the model invents a package or label: keep the exact reference, remove extra props, or return to the master photo. Leave one checkable artifact from this step; do not start the next until it exists. 5) Test at the lowest cost: A 5-second Agnes Video test is 10 credits. Upgrade only the winner to a paid render: 90 credits for 5 seconds of Seedance 480p or 195 credits for 720p. Leave one checkable artifact from this step; do not start the next until it exists.
Change one named variable per rerun: prompt, reference, camera, duration, model, ratio, or export crop. If two variables change together, a better output cannot be reused because nobody knows which fix worked.
Review gates and evidence
Review at 100%. Read small text, count buttons and parts, compare color against the source, inspect the four corners, and overlay the platform crop. A thumbnail is only a concept check, not a marketplace check.
Keep source, approved wording, rejected version, correction, credit cost, and final crop next to the asset. The record exists so the next reviewer can reproduce the decision without a chat thread.
Failure diagnosis and retry ladder
Name the failure first: label redraw, color drift, shape drift, missing accessory, wrong contact shadow, or cropped CTA. Stop prompting for more letters, reduce the variable count, return to the reference, and only then upgrade the model.
Do not upgrade on hope. A 5-second Agnes Video probe is 10 credits. If the concept passes, Seedance 2.5 costs 90 credits at 480p or 195 at 720p for 5 seconds; Kling is 195 without audio or 245 with audio. Veo enters only when a 4/6/8-second cinematic bucket is genuinely worth 80/125/165 credits. Prove hook, subject, and rhythm first, then pay to clean up motion that already worked.
Versioning and handoff
Deliver by module and ratio, not one stretched master. Name files with SKU, module, crop, version, and approval state. Keep the rejected render and the correction line in the same folder.
The folder is boring on purpose: source, approved reference, generation settings, caption file, platform cuts, QA screenshots, and one correction line. A reusable asset is boring in the right way.
Why this is not a generic answer
The imgmov advantage is the chain: Asset Library preserves the subject reference, Canvas fixes shot order and first/last frames, the workspace routes Agnes to Seedance, Kling, or Veo only after proof, and your external editor adds captions and CTA without regenerating the media.
Do not invent certifications, competitor claims, price windows, or a model using the product in an unsafe way. If the seller cannot prove it, the image does not carry it. If the request only asks what ai product video prompts means, a search page is faster. This page is useful when someone must deliver a reusable prompt and a matched model setting under real constraints.
Frequently asked questions
What is the safest prompt structure?
Protected product, camera, one variable, duration, and correction. Do not bury the protected fields in a long style sentence.
The workflow is specific about its stopping point: it starts with product attributes, platform, and desired action and stops at a reusable prompt and a matched model setting. In imgmov, upload the source, lock it as a reference, set 9:16, 1:1, or 16:9, run one proof, and save the passing settings. If a reviewer cannot tell which version is approved, the workflow has failed even when the media looks good.
Can I ask for price text?
No. Add price and claims in the external editor so a copy change does not force a rerender.
Do not upgrade on hope. A 5-second Agnes Video probe is 10 credits. If the concept passes, Seedance 2.5 costs 90 credits at 480p or 195 at 720p for 5 seconds; Kling is 195 without audio or 245 with audio. Veo enters only when a 4/6/8-second cinematic bucket is genuinely worth 80/125/165 credits. Prove hook, subject, and rhythm first, then pay to clean up motion that already worked.
How many prompts should I test?
Two for the same SKU: one clean use scene and one detail motion. More variants waste credits before the reference is proven.
The check is not “does it look AI-nice?” It is: Review at 100%. Read small text, count buttons and parts, compare color against the source, inspect the four corners, and overlay the platform crop. A thumbnail is only a concept check, not a marketplace check. Then the source, approved copy, rejected version, correction, and final crop stay in the same handoff folder.
What makes a label blur?
Fast camera motion, redrawing text, and too many props. Shorten the motion and forbid text redraw.
Name the failure first: label redraw, color drift, shape drift, missing accessory, wrong contact shadow, or cropped CTA. Stop prompting for more letters, reduce the variable count, return to the reference, and only then upgrade the model. Change one named variable, keep the old version, and record the credit cost. If the same failure repeats, fix the reference or scope rather than asking the prompt for forgiveness.
Which model should prompt templates target?
Draft with Agnes Video, then move the winner to Seedance, Kling, or Veo. Kling snaps to 5 or 10 seconds; Veo snaps to 4, 6, or 8 seconds.
The imgmov advantage is the chain: Asset Library preserves the subject reference, Canvas fixes shot order and first/last frames, the workspace routes Agnes to Seedance, Kling, or Veo only after proof, and your external editor adds captions and CTA without regenerating the media. The handoff is reusable only when the source, reference ID, prompt, model, cost, rejection reason, and approved cut are linked. That chain is what makes the next ai product video prompts task faster.