All workDiagram only: the client's work is under NDA.
From a 3D model and a prompt to a batch of marketing images.
- What
- AI render pipeline
- For
- A product-visualisation company (client)
- Built with
- Blender, Python, cloud GPU
- Status
- Delivered
- 200+ product models3D model+ a text prompt
- Scene set by an LLMcamera, light, placement
- Cloud GPU renderserverless, batched
- 1024 × 1024Marketing images4 per product
The problem
Studio-quality product images took manual setup for every product and every angle.
What it was
Blender tooling and a cloud GPU backend that turn 3D models into marketing images.
How it was done
- Blender add-ons that set up scene, lighting and camera automatically.
- A serverless GPU backend that takes a model and a prompt and returns a batch of renders.
- LLMs choose lighting, camera and placement, with fallbacks so a render never fails.
The result
Four renders per product at 1024 × 1024, over a library of 200+ product models.