CASE SUMMARY
Content automation is not one-click publishing. It shortens the repetitive work between source material and an approvable asset.
The client is an insurance and financial planning team serving Hong Kong. Every topic required research, Traditional Chinese rewriting, five-slide structuring, brand checks, image export and caption writing.
The new platform accepts websites, YouTube, PDFs, DOCX, text and CSV as references, then generates an editable carousel. AI handles preparation and the first draft, while people familiar with the business and compliance review before publication.
Research, copy, design and export lived in separate tools, so one change triggered a full round of rechecking.
Sources included news sites, regulatory documents, product PDFs and videos. A content specialist read the material and planned a cover, three content slides and a call to action, then a designer handled Chinese typography, wrapping, logos and image ratios.
Long copy could overflow, while browser previews and downloaded images could shift because of font metrics. A late compliance or CTA change often required a full re-export and slide-by-slide review.
Source material is first structured, then turned into five slides with defined roles.
The system extracts readable website or document content, using direct text for suitable PDFs and image understanding for scans. Reference material is separated from system instructions, and source names remain visible in the editor.
A standard carousel contains a cover, three content slides and a CTA. Generated output passes structural checks, with missing fields repaired or filled using safe defaults before the template receives the data.
Four templates share one data model, with key copy and brand elements editable in real time.
Classic, Magazine, Portrait and Quote templates can be switched without regenerating content. Titles, body copy, step labels, brand name, Instagram handle, disclaimer and CTA are separate fields.
Logos preserve their full ratio, while portraits and cover images can be cropped, moved and scaled. HEIC or HEIF files are converted to a web-compatible format, and the original remains available for later adjustment.
Automatic fitting solves overflow, while a canonical PNG keeps preview and download consistent.
Each slide uses a fixed 1080 by 1350 frame. When text is too tall or wide, a binary search finds the largest fitting scale and handles URLs or English strings without natural break points.
Before capture, the system waits for fonts, images and fitting to finish, then locks the measured text widths. Preview, individual download and ZIP export reuse the same PNG artefacts, reducing post-download layout surprises.
The client initially estimated an 80% improvement and about 60 staff hours saved each month.
The team now handles source preparation, copy, layout, branding and download in one interface. Version history also reduces accidental overwrite and rework, freeing time for fact checks, strategy and pre-publication review.
The figures remain early client estimates based on current workload. Formal measurement should cover four to eight weeks of production time per set, rework, output and team participation.
Frequently asked questions
Can every line be edited after generation?
Yes. Main titles, body text, labels, CTAs, brand fields and disclaimers remain editable rather than being baked into an image.
Does the system publish automatically?
The focus is producing approvable assets. Scheduled generations still enter a review state for team approval before publishing.
Can long Chinese text overflow?
A shared text-fitting component measures and scales the text group with a minimum ratio. Extremely long content should still be shortened by an editor.
EXPERT AUTOMATION CONSULTATION
Does your team have a repetitive, error-prone workflow?
Tell us how the process works today, its monthly volume and the common exceptions. We will identify the part worth automating and define a testable first phase.