Tindra Market
A marketplace standardised inconsistent seller photography into one clean catalogue look through the API.
Fashion + home marketplace, 60,000+ listings, thousands of sellers
Catalogue standard
Seller uploadTindra's sellers uploaded photos in every imaginable lighting, crop and background. The catalogue looked like thousands of different shops, because it was.
Inconsistent imagery drove returns: buyers received items that looked nothing like the washed-out or over-edited listing photo.
Manual moderation of image quality did not scale past a few thousand listings.
Every seller upload passes through a Vela normalisation step before it goes live: background standardised, lighting relit, resolution lifted.
Sellers get the same clean output through a self-serve API, so the standard is enforced at ingest rather than policed after the fact.
Outputs carry C2PA provenance metadata, so the marketplace can prove how each listing image was processed.
The exact endpoint chain.
Composable primitives, run in order on a webhook. No glue code beyond the calls themselves.
- 1Standardise background
POST /background/change/image/v1One clean catalogue backdrop across every seller.
- 2Normalise lighting
POST /edit/image/v1Relight mixed-source photos to a single look.
- 3Lift resolution
POST /upscale/image/v1Bring low-res seller uploads up to catalogue spec.
Over 60,000 live listings were normalised to one catalogue standard, with new uploads conformed automatically at ingest.
Returns attributed to listing mismatch fell 31% as photos finally matched what shipped.
Median processing landed under 900ms per image, fast enough to run inline on upload.
We turned thousands of different shops into one catalogue without telling a single seller to re-shoot. It just happens at upload.
Your catalogue, your numbers.
Start free with 250 generations and benchmark it against your last shoot.