Bria Embed Product

Bria Embed Product is best suited to placing approved product sources into prepared lifestyle, campaign, collection, or marketplace scenes. It is especially useful when a merchant already knows the desired location and size and wants structured coordinate control rather than prompt-only composition.

The current Masonry contract requires one background under `first_frame_image` and an `elements` array with at least one product. Every element needs an image plus integer x, y, width, and height coordinates. Width and height must be greater than zero, while x and y cannot be negative.

The live `bria-embed-product` route checked August 17, 2026 accepts 1–10 product elements. With the required background, that is up to 11 input images. A multi-product request still needs a separate box and truth review for every item.

Its meaningful controls are the background image, product elements, coordinate boxes, and optional seed. Masonry's multimedia CLI currently requires a prompt argument at submission, so a neutral description can be supplied, but placement semantics come from the structured input rather than prose. Do not expect a text prompt to replace missing elements or coordinates.

Run `masonry multimedia params bria-embed-product`. Use the multimedia command because the simpler image-parameter view does not express this route's required `elements` array. Recheck the live contract before turning a sample into a durable batch script.

Submit `first_frame_image` plus `elements` in the multimedia input JSON. A single-product element can look like `{"image":"https://example.com/product.png","coordinates":{"x":150,"y":80,"width":300,"height":560}}`. Replace both URLs and all four box values with the dimensions of your own approved files; the example is a schema illustration, not a universal placement.

A controlled source and narrow task reduce ambiguity, but the output is still generative. Masonry's retained bottle sample kept all visible identity checks used in that run, yet one success cannot establish repeatability or pixel identity. Compare each output with approved source files and reject any changed geometry, label, color, quantity, variant, claim, or included item.

It is better matched when the merchant already has both the product and scene and needs explicit placement. A general model can be more flexible for inventing an environment, changing art direction, or following a complex prose brief. The right comparison is accepted output per job, not which model made the most dramatic image.

Yes, the route accepts up to 10 products, so it can assemble a bundle candidate from separate approved sources. Verify relative scale, count, overlap, front-to-back order, shadows, packaging, and whether every visible item is actually included in the offer. A beautiful arrangement must not imply a different bundle.

Measure the desired box against the background image's pixel dimensions. Start with one product, keep the box inside a physically plausible surface or open region, and review the generated scale against scene cues. Save the exact x, y, width, and height with the output so a layout decision can be reproduced or revised deliberately.

Start from the highest-quality product image you are licensed to use, preferably a front or three-quarter view that matches the destination camera angle. Keep the original file, approved label artwork, SKU and variant ID, physical dimensions, and visible-truth checklist together. A clean edge or uncluttered source gives the model a clearer product signal, but do not remove a real shadow or reflection if it contains material information you need to review. Prepare the background separately at its intended aspect ratio, leave physically plausible room for the product box, and avoid scene cues that imply an impossible scale. For a batch, version every source instead of silently replacing files under the same URL.

Masonry exposes an optional integer seed and currently defaults it to 5555. Record that seed with the exact source files and boxes. A seed improves the audit trail, but it does not guarantee identical results after a hosted model or processing pipeline changes.

Not by default. Keep approved factual photography as the primary record unless the output passes the channel's current rules and a strict exact-SKU review. Generated contextual scenes are usually safer as secondary media, ads, collection art, or layout candidates; deterministic compositing is preferable when source pixels must not change.

Track submitted candidates, service failures, product-truth passes, scene-quality passes, review minutes, correction minutes, and cost per accepted asset. Downstream CTR or conversion should be measured only after the creative is approved and exposed to enough traffic. Do not report downloads as completed assets when review rejects them.

One controlled run placed a fictional sage NORTHLINE bottle into a supplied neutral scene using x 150, y 80, width 300, and height 560 with seed 73017. The retained output preserved the visible silhouette, cap and loop, color, wordmark, and 750 ML line while adding a plausible shadow. That observation applies to one output and is not a model-wide fidelity claim.

Bria Embed Product is an AI image generation model from Bria, available inside Masonry, the AI creative agent teams use to produce marketing, product, and brand images.

Open a Masonry canvas, pick Bria Embed Product from the model selector, and describe the image you need: a product shot, an ad creative, a social post. Masonry generates it, then you refine, edit, and combine Bria Embed Product with other models in one workspace.

Yes, you can start generating images with Bria Embed Product on Masonry's free tier, then scale up with higher limits and priority processing as your team grows.

Treat the scene and coordinate boxes as the brief. Start with a clean, licensed background whose perspective and available surface make physical sense. Use a product image with a clean edge, define the box in the background's pixel coordinate system, and run one placement before scaling a batch. Review silhouette, closure, color, label artwork, exact words, quantity, relative size, occlusion, perspective, contact shadow, and unsupported additions. If the sold SKU must remain pixel-identical, generate or select the empty scene and composite the approved product deterministically.

Bria Embed Product is built by Bria. Inside Masonry it runs alongside 50+ image and video models, so your team can pick the right one for each brief without switching tools.