Every tile dealer wants more foot traffic
Lao Zhang has been in the tile business in Chengdu for 8 years. A 200-square-meter showroom, mostly residential clients. Last year he started worrying — foot traffic was drying up, renovation companies were eating most orders, walk-in customers came only through referrals.
He tried posting videos on Douyin and Xiaohongshu. Filmed tile displays with his phone, slapped on some text. "Terrible results," Zhang said. "Two hundred views in two weeks. Zero inquiries."
This March, he hired a young sales rep named Xiao Chen. Xiao Chen said let me handle the social media, but we need AI.
What they actually did with AI for short videos
Xiao Chen built a workflow around one idea — let AI handle the heavy lifting.
Step 1: ChatGPT writes scripts in 10 minutes
The old approach: Zhang would point at tiles and say "this quality is good, this color is nice." Dry, boring. Xiao Chen had ChatGPT rewrite the whole approach — sell the scenario, not the product.
Every script revolves around a specific spatial scenario: "How to budget a bathroom reno under 30K RMB?" "Pitfalls to avoid when re-tiling an old apartment." "How a young couple picked tiles for their 120m² wedding home." ChatGPT generates the script outline. Xiao Chen polishes it in 10 minutes into conversational Chinese.
Zhang's words: "Before, what I said put customers to sleep. AI organized it, and suddenly I sounded like a renovation influencer."
Step 2: Kling AI generates the visuals — no camera needed
The real headache was footage. Tile photography is finicky — wrong light skews colors, wrong angle distorts textures. Hiring a photographer used to cost 2,000 RMB per shoot.
Xiao Chen switched to Kling AI's text-to-video. Input: "Modern minimalist style, gray tiles, 60×120 large format full-floor layout, living room natural light, wide angle." Thirty seconds later, a 15-second showcase video. The lighting, angles, and texture all looked better than phone footage.
Silent B-roll with voiceover wasn't enough. They added Jianying AI's digital human feature — record Xiao Chen reading a script once, AI learns the voice and syncs mouth, pace, and expressions to any new script. No more re-recording for every video.
Step 3: Jianying AI auto-edits — 3 videos per day
Before, editing one video took 2 hours. Now Jianying AI auto-adds subtitles, background music, and color grading. Xiao Chen spends 30 minutes a day producing 3 videos, posting to both Douyin and Xiaohongshu simultaneously.
Two hard lessons
Lesson 1: Same content on both platforms bombed hard. Xiao Chen initially cross-posted Douyin videos directly to Xiaohongshu. Views tanked. Turns out the content logic is completely different — Douyin hooks in the first 3 seconds, Xiaohongshu hooks with the cover image. After adjusting, Douyin used "question first, then answer" pacing. Xiaohongshu used comparison images plus checklist format.
Lesson 2: AI-generated visuals looked too fake. Kling's tile renders often felt disconnected from reality. Xiao Chen found a trick — first take a close-up photo of the real tile with a phone, then use Kling's image-to-video mode. AI anchors on the real texture and only generates the environment around it. Much more realistic.
Numbers after 3 months
- Douyin/Xiaohongshu followers: 0 → 8,000+ / 6,000+
- Daily inquiries: 0 → 32 (23 Douyin + 9 Xiaohongshu)
- Monthly closed deals: ~15 (previously ~5-6 from storefront + referrals)
- Cost per video: ~4 RMB (split Kling + Jianying subscriptions). Total monthly spend: ~360 RMB
- Output: 3 videos/day, ~90 videos/month
Zhang's favorite part of the day now is opening the Douyin backend to check incoming messages. "Before, I'd get maybe two calls a day. Now my phone buzzes constantly with potential customers. This 360 RMB a month beats local renovation platform ads by a hundred times."
Zhang is thinking about two next steps. First, replicating from one category (tiles) to multiple categories — same tools and workflow, just swap Kling AI prompts from 'gray tile series' to 'walnut flooring' or 'sintered stone countertops.' In theory, bathroom fixtures, flooring, and coatings can all work with the same playbook. Second, he is considering hiring a dedicated short-video operator to raise daily output from 3 to 6 videos across more accounts. 'The bottleneck isn't the tools — it's the people. AI has pushed content production cost near zero, but topic selection, script direction, and comment engagement still need human input.' Nobody buys tiles from one video. But short videos solve the first step of lead gen — making strangers aware you exist. When AI pushes content production cost near zero, customer acquisition shifts from "word of mouth" to "continuous exposure."
Zhang put it bluntly: "I don't understand how AI works. But I understand business — if 30 people ask me about tiles every day, the sale is just a matter of time."