A Web Shop is a showcase and catalog with prices, SKU rules, promotions, and integrations with billing/analytics. Previously, a "live" result required assets, texts, catalog structure, prices, and numerous edits. For partners, this is a high barrier to entry: they want to quickly see what their store will look like without having to gather materials and involve their team.

In the first versions, we used parsing, which automatically pulled content from the game page on the platform. The "wow" effect was there, but the preview was not accurate: artifacts remained in the texts, products were missing, and visual consistency was broken. We needed a way to deliver fast and predictable results without manual preparation so that the partner would have an "aha moment" at the start of integration.

Task

Design a process for generating store content using AI: texts and images for a specific game/partner. Requirements: fast, predictable, repeatable; with control points and the ability to edit manually.

Actions

1. Flow in Figma.
   URL → context parsing → text templates → asset generation → showcase assembly. Fixed quality criteria and manual editing points
   
   

   

   
   

2. Fail #1. Plan with Midjourney.
   Designed a flow based on reference images in MJ, agreed with the lawyers on the ToU — but after review with development, we ran into the lack of an API for MJ and SLA for providers; integration is not possible.
   
   

3. Fail #2. Switching to Stable Diffusion.
   It was the only option with API and SLA/hosting that supported references. The quality/consistency fell short of requirements.

   

   
   

4. Rebuild.
   Assembled an independent n8n pipeline: first texts/prompts, then images; created quick scripts for standalone tests and provider comparisons, without involving development.
   
   

   

   
   

5. Choosing a working alternative.
   I tested several models/services and recorded the parameters. The current stack provides stable styling and repeatable results, and importantly, it is now much easier to test new approaches and changes through n8n.
   
   

   

   
   
   

6. Transfer, testing, and polishing.
   Described the API/events, handed it over to development; testing on several titles. And the n8n flow became an example of feature architecture.

Result

The showcase is assembled in approximately one minute from a single link; visuals and texts correspond to the style of the game, and the result is repeatable; the entry threshold for partners has been lowered an "honest preview" appears without lengthy manual preparation.

Lessons

• API и SLA it is just as important to check the legal aspects before designing a solution.

• Meaning → images: separating text and images gives you control over quality; this really boosted the result;

• Fast iterations via n8n must-have: you can test hypotheses without involving development.