We took a step back and rethought our whole pipeline around the recent changes. Our product is megascans, so the goal was simple: maximize output with minimal manual work.
First decision — we completely dropped the subscription model. It just doesn’t fit how we operate.
Then I built a Blender addon that automates almost everything:
* generates previews from all angles
* creates clean LODs down to LOD6 without texture loss
* handles export (formats, mesh, etc.)
One click, ~7.5 minutes — and the asset is fully prepared.
On top of that, I made an AI assistant. It reads the addon’s log file, extracts all relevant data, and generates title, description, and keywords. You just type one word (e.g. “banana”), and it does the rest. At that point, it’s just copy-paste into the platform.
So total upload time per asset (including previews + metadata) is about 8 minutes, almost fully automated.
For PBR textures, we use a separate tool: you input the color map, and it generates the rest. That takes ~2 minutes.
The scan creation itself is also largely automated. With our current setup (even on older hardware), we get about 1 finished megascan per ~40 minutes of compute time, with minimal human involvement. Roughly speaking: ~1 scan per hour, passively.
Based on this, we decided to segment our content by value:
* simple, easily obtainable assets (rocks, walls, stumps, etc.) → mass-upload to this platform
* rare or high-value assets (antiques, unique props, etc.) → distributed to other marketplaces
So instead of treating everything equally, we now split assets into “buckets” depending on their value and rarity. This approach makes the whole system much more efficient and scalable.
