The need
Running an AI content factory (Gemini + n8n) for restaurant menu posts, his generated videos don't present food the way Thai promos need, and AI-added price text (e.g., 129 baht promos) never comes out right or appetizing.
Original post — verbatim Thai
ผมอยากจะใส่ราคาลงในภาพที่ผมถ่ายโดยให้ AI เป็นคนใส่ให้อารมณ์คนเห็นคอนเท้นแล้วอยากกินนะครัย
Three business opportunities
OPPORTUNITY 01
Thai menu video QA guard
- What
- Pipeline stage (API/n8n node) that verifies generated promo frames against the menu data — exact price text, dish-photo fidelity — and regenerates until passing.
- Who pays
- Content factories pay per clip because a wrong price in a posted promo is a refund disaster.
- Price
- THB1.50/clip (hypothesis)
- Why now
- He already wired Gemini+n8n and is stuck on exactly this failure; vision-LM checkers became cheap in 2026.
- Effort
- M
- Moat
- Thai F&B visual QA dataset: which frames make food look appetizing, labelled over thousands of runs.
- Risk
- Newer video models render text correctly and the guard's job shrinks.
OPPORTUNITY 02
Deterministic promo overlay studio
- What
- AI proposes composition; the price/promo text is rendered deterministically on top with Thai fonts and appetite-optimized styling — zero hallucinated numbers.
- Who pays
- Restaurants and freelancers subscribe per store to kill the exact pain of AI-written wrong prices.
- Price
- THB299/mo/store (hypothesis)
- Why now
- Current models still fumble on-screen Thai text; sidestepping generation with overlays is a 2026 no-brainer.
- Effort
- S
- Moat
- Style library of overlay designs measured by engagement per cuisine type.
- Risk
- Canva-class tools bundle similar templating.
OPPORTUNITY 03
One-restaurant content factory service
- What
- Done-for-you tier: shop uploads dish photos weekly, receives scheduled FB reels with correct promos; internal tooling hides AI flakiness from the client.
- Who pays
- Single-store owners pay a flat monthly fee because hiring an agency costs more and DIY failed them.
- Price
- THB4,900/mo/store (hypothesis)
- Why now
- This post proves SMBs will DIY with AI and fail; service arbitrage with tooling in the middle is the gap.
- Effort
- M
- Moat
- Local portfolio of appetite-raising Thai food reels compounds per cuisine vertical.
- Risk
- Unit price strains Thai SMB budgets; churn when results plateau.