DapurStudio

DapurStudio turns one phone photo of a dish, taken in a messy seller's kitchen, into a studio-style catalog photo, with an ingredient cost range and an Indonesian sales caption. The seller uploads a photo. SAM 3.1 segments the dish, a pre-validated studio backdrop template is reused (never generated per photo), and the cut is recomposed in studio light. Muse Spark identifies the dish, lists the ingredients, estimates a cost range, and writes a sales caption. The server picks the safe cut automatically behind the scenes. A review status, separate from pipeline success, plus plain-language advice tell the seller when a reshoot is better. The seller edits the dish name, price range, and caption, then saves the card to the catalog. Built during the hackathon with Meta AI tooling: SAM 3.1, Muse Spark, and Muse Image (backdrop template only, validated once and reused). Honest limits: dish naming varies between runs, so the seller always reviews before saving. The content gate is best-effort, not a certified moderation filter. No impact numbers are claimed, no real seller has measured sales impact yet.