AI built for African markets
Our in-house AI work focuses on computer vision. We practise ethical AI: we collect and own our data, and we build for the people existing datasets leave out.
ThelerineAI
Visual intelligence for fashion. Virtual try-on, built from the ground up.
A computer-vision pipeline that lifts a garment out of one photograph and fits it to another. Geometry first, learned refinement next. Built in South Africa, for the bodies the fashion industry usually skips.
Thelerine is the word Batswana and Basotho use for a skirt. It probably comes from Terylene, the fabric.
How it works
Four steps from photo to fit
- 01SegmentLift the garment out of a worn photo, not only studio flat-lays.
- 02FlattenRecover the garment's shape so it can be placed on a new body.
- 03FitWarp it to the target pose with pose-anchored geometry.
- 04BlendComposite it into the target photo.
Garment extraction
Promptable segmentation lifts garments from real photos.
Geometric fitting
Nothing is invented, so nothing can be hallucinated.
Learned refinement
A network to resolve seams, shadow and fabric behaviour.
Representative data
Bodies and skin tones that existing try-on datasets under-represent. We don't scrape images from the internet.
Roadmap
Where ThelerineAI is going
Now
A small, owned dataset
It works on the images it was developed with, not yet on any photo you give it.
Next
A retail partner
Learning from a shop's catalogue, with consent.
Later
Size recommendation
A model that tells you which size to order.
“Virtual try-on should work for everyone. Built by Africans, for Africans.”
From the blog · 31 January 2026
ThelerineAI: the journey
Notes from building our own dataset to train the virtual try-on model.
Read the post