Every store doing AR eventually hits this fork: pay a 3D asset production SaaS pipeline per model, or build the capability in-house. There is no universal answer, so here is the honest comparison.
SaaS wins on speed and simplicity. Upload photos or scans, get back models that are cleaned, converted, and store-ready. No hiring, no software licenses, no learning curve. For catalogs under a few hundred SKUs, or a first AR experiment, this is almost always the right call. The per-model cost is visible, the quality is someone else's problem, and you can start this week.
In-house wins on volume and control. If you are adding hundreds of products a month, per-model pricing starts to hurt. A small capture setup with a seller phone-scan app for 3D assets plus one trained person can pay for itself fast. You also control turnaround. New product Monday, 3D model Tuesday, no vendor queue.
The hybrid is underrated: SaaS for the backlog, in-house for the ongoing flow. Clear the existing catalog with a vendor, then bring new-product capture inside once the process is proven.
Whatever you choose, keep file hygiene from day one. Versioned models, clear naming, one source of truth. Asset libraries get messy fast, and a messy library quietly doubles every future cost.
The mistake to avoid is building in-house before you have proof AR works for your products. Rent the capability, prove the behavior, then buy the capability. In that order, always.
Figuring out which route fits your catalog and volume? I consult e-commerce brands on AR implementation at arcommerce.fyi.