Start with the missing work
Yinghanni is a precision manufacturer. At the time, it lacked a dependable process for producing English content. The website needed clearer product and purchasing information for overseas buyers; company videos also needed English versions, followed by processing and publication.
Without a repeatable workflow, each request became a one-off: find a translator, process the file again, and publish it manually.
Build the website first, then connect video
I began with the website and buyer-facing information so that buyers had a stable place to check the company and its products.
HeyGen handled English lip-synced versions of existing videos. n8n connected the later processing and publishing steps. Each tool had a defined job; product claims and final publication still remained under human control.
What people still had to confirm
Before automation, the company needed to decide:
- how product names and specifications could be described publicly;
- which images, cases and customer information were publishable;
- what purchasing information mattered most;
- who made the final check before release.
I linked the content plan, website, video workflow, and publishing steps. The company remained responsible for product facts. Automation was limited to repeatable steps with clear rules.
Why we did not start with a tool
The first decision was to identify the work that could not be repeated reliably. In this case, that meant English content that required a new translator, another round of processing, and manual publication each time.
I still follow the same order: find work that cannot be done consistently, then choose the tool. Starting with the tool often produces a convincing demo that no one maintains.
Current state and boundary
Confirmed: the English website is public. HeyGen was used to produce lip-synced English versions of existing videos, and n8n was used to connect video processing and publishing steps. Final publication remained subject to human review. A saved Search Console snapshot from August 26, 2026, showed 22 canonical pages indexed at that time.
Boundary: that number is a dated search snapshot. It does not prove sustained rankings, inquiries, customers, or orders.
Public site: Yinghanni. Related method: what I check before a business website goes live.