What this case covers
The author has been doing cross-border e-commerce in Yiwu for 8 years, organizing product images, benchmark videos, AI image generation, video generation, and post-production editing into a reusable sales video workflow, packaged as a skill. The article covers the complete process from understanding the product to final video acceptance.
Questions it answers
- How do you package an AI sales video workflow into a skill?
- How do you solve product distortion and empty shots?
- How do you handle foreign language dubbing and language issues?
- How do you get AI to understand the product and output a storyboard?
- How do you consolidate and reuse a finished video?
Our observations
- The value of a reusable skill lies in turning one-off production into a repeatable, callable asset
- Manual review of the first frame is the key checkpoint for controlling product distortion and visual quality
- Using a cheap video generation platform for batch output is suitable for volume runs rather than pursuing a single polished piece
From the author
I use Codex to organize materials, write plans and prompts, and call codex tools to generate images and Thai audio; video clips are generated through the image-to-video platform Google Flow, and finally processed in CapCut for subtitles, sound, and pacing. Finally packaged into a reusable skill