Codex + Skill + Infinite Canvas: Run a Full E-Commerce Content Pack from Two Images

Course intro

From a massage-gun promo poster to a complete product content pack.

1. What the Skill does

This Skill turns a massage-gun promo poster and a fitness-outfit reference into a complete content pack ready for e-commerce display and video generation. It uniformly extracts product structure, brand identity, core selling points, character outfit and usage scenes, then continuously produces white-background product images, Chinese/English detail pages, platform posters, campaign posters, character assets, scene key visuals and vertical selling videos.

It specifically solves: product structure drifting across images/videos, wrong character-grip relations, scene-product mismatch, and fragmented assets across platforms.

2. What to prepare

- At least one promo poster or product image showing the massage gun's structure, color, brand and main selling points.

- If a character usage scene is needed, prepare a target fitness-outfit reference.

- Optionally provide brand model, real specs, price, campaign mechanic, target platform, aspect ratio and banned expressions.

- Without full product specs, you can enable online auto-complete mode; the Skill fills production params from public info and image clues.

3. How it works

1. First extract the massage gun from the promo image and generate a clean white-background product asset.

2. Using the white-background product image as the unified product input, generate Chinese detail page, Taobao poster, 618 campaign poster and English cross-border detail page.

3. From the outfit reference, build an adult female fitness model, yoga mat and character three-view, then complete the massage-gun grip and calf-contact relation.

4. Generate a seamless yoga-room key visual, locking character, product, scene, motion and lighting.

5. Form a 3:4, 12-second, 720p three-shot execution plan and output the final effect video.

4. Two usage modes

User-asset mode: use the user-provided product poster, outfit image, specs, price and campaign requirements, generating step by step after node confirmation. User info always takes priority; only after all key images are confirmed does it move to the next stage.

Online auto-complete mode: when brand model, product specs, price, selling points, clothing style or platform params are missing, the Skill matches public info online from image clues and reasons to fill needed params without breaking the production chain.

5. Boundaries to stress

- Online-filled info suits simulating the full production flow; before real launch, defer to the actual product page, stock, price, promo policy and platform compliance.

- Accounts, verification codes, paid credentials, internal brand data, privacy info and asset licenses cannot be auto-filled or done on your behalf.

- A single product image cannot prove occluded structure and all specs; generated content must be based on visible evidence and confirmed info.

- Where real-person likeness, brand marks, design copyright or third-party assets are involved, the user must confirm the legal license scope.

- Final quality still depends on chosen image/video models, account quota and platform capability; obvious product distortion, wrong grip, character deformation or product occlusion must be reworked.