We build AI product photography that ships in under 15 seconds.

What is Palmou AI?

Palmou AI is a web-based product-photography studio for e-commerce sellers. It does two things, chained inside one credit: it removes the background from your product photo cleanly and generates a new scene around the product. You upload one phone photo, optionally pick one of six predefined scenes, and Palmou returns a catalog-ready WebP image in under 15 seconds. Each completed image costs 1 credit. There is no app to install, no Photoshop skill required, and no per-feature paywall — the price you see is what you pay.

We're focused on a single workflow: one source photo → one finished catalog image. We don't do video, avatars, or generic image generation. The narrowness is intentional; it's how we keep the pipeline calibrated for the highest possible output fidelity per credit.

Who built it

Palmou AI is built by Palmou Studio, an independent software studio based in Spain. The product launched in 2026 after several months testing the new generation of vision models that became commercially available that year. We operate as an EU-based company, subject to GDPR and the Spanish LOPDGDD.

You can reach the team at palmoustudio@gmail.com or on X at @PalmouAi.

Why now: the 2026 model landscape

Until 2025, AI product photography stacks needed to manually paste cutouts into pre-rendered scenes. Two releases changed the picture:

• BRIA RMBG 2.0 — a background remover with a 256-level alpha channel instead of a binary mask. The difference is clearest on hair, fur, glass, jewelry chains, and any reflective product where the old generation visibly halos.
• Google Gemini 2.5 Flash Image — image-to- image scene generation that preserves product identity without requiring a precomputed mask. Replaces the older SDXL + ControlNet inpainting workflow.

Palmou AI's core insight is that the two best models for the two steps of a product-photo pipeline now exist — they just live on different platforms. Our job is to chain them with the right defaults, hide the complexity, and charge a flat 1 credit per image.

Tech stack

Palmou AI runs on Next.js 16 (App Router) and React 19, hosted on Vercel. Auth, database, real-time, and storage are all on Supabase in the EU region. Payments go through Stripe Checkout and the Stripe Customer Portal. Both AI models run on Replicate; we keep webhook signatures verified so forged callbacks can't inject results. Errors flow into Sentry. Transactional email is Resend.

We picked Replicate over self-hosting GPUs because at our scale the inference cost is much lower per generation, and the latency is competitive after the first warm request. When we outgrow that trade-off we'll move specific models in-house.

How we think about pricing

One credit = one finished image. No per-feature paywall, no watermark on any plan. Reprocessing the same image with a different scene costs 0.5 credits because most of the cost has already been paid. Failed jobs (content moderation, model timeout, Replicate outage) auto-refund. There's no free trial on paid plans because we don't want to incentivize disposable- email signups; we'd rather give you 1 free credit on signup with no card and let the output speak for itself.

Privacy and ownership

Your uploaded images stay in Supabase Storage in the EU, encrypted at rest. They are never used to train models — not ours and not third parties'. The only network call that leaves Supabase is the inference request to Replicate; the output comes straight back into your namespace and can be deleted at any time. You own every image you generate and can use it commercially without restriction.

What's next

Roadmap items we've committed to publicly: a public REST API (already shipped on the Agency plan, see docs/api), automated batch processing for marketplaces, and more scenes voted on by the community feed. We don't plan to add video, avatar generation, or text-to-image — there are better tools for those.