PixelOkay Images stay on your device
Tools
CREATIVE IMAGE STUDIO FLAGSHIP

AI Product Scene Generator

Transform product photos into polished studio shots and lifestyle compositions. Build scenes, refine a cutout, perfect the light, and export marketplace-ready images—right in your browser.

✦ Optional local AI cutout◎ Custom backgrounds↓ High-resolution exports

YOUR CREATIVE WORKSPACE

Product Scene Studio

LIVE COMPOSITION

Softbox studio

● Live preview

Bring your product into focus.

Upload a photo or open the example. Your scene will appear here.

04 / EXPORT STUDIO

Ready for your storefront.

PNG for quality, JPEG for smaller files. Build multiple looks in a single download.

Adds a lower-third to the finished scene (not the transparent product cutout). Avoid including personal information.

Select scenes for batch ZIP (six checked by default)

Full Project ZIP includes the product, edited cutout, custom background and settings. Scene-settings JSON contains styles only, not private images.

Exports are generated on your device. Scene presets are procedural illustrations, not AI diffusion photographs.

LOCAL AI SETUP

Set up AI on this device

First, a model needs to be loaded on your device

ORMBG removes backgrounds from product images locally with Transformers.js and WASM. Model and runtime assets download only after you explicitly select Download & Load. Your product images are not uploaded for model inference.

  1. 1. Check device
  2. 2. Download / cache
  3. 3. Set up AI
  4. 4. Ready

ORMBG · onnx-community/ormbg-ONNX · q8 · WASM. It creates a subject cutout for editing. No photographic scene synthesis is claimed. Model RAM requirements are unverified; WebLLM text models are not compatible.

Device & model resources

Browser-reported estimates
Approximate device RAMChecking…

Live model RAM use cannot be measured in most browsers.

Browser website storageChecking…

Estimated site storage, not RAM.

ONE ACTION · ALL RECOGNIZED AI MODEL DOWNLOADS

Free AI model storage across PixelOkay

Scan this website origin for WebLLM and Transformers.js model caches, plus recognizable model artifacts in other cache buckets. This includes caches from other PixelOkay tools. No site data or browser databases are cleared automatically.

Scan to inspect recognizable model caches. Unknown data is skipped.

Unknown IndexedDB, OPFS, browser-managed models and app data remain. Close other AI tabs before cleanup.

Having trouble downloading or loading AI?

Try Refresh Storage & Cache, retry loading, or continue using solid-bg cutout. Only consider model-only cache removal after inspecting the targets. Clearing all website data can delete your projects and drafts.

AI is optional. Solid-background removal and scene export remain available.
Currently loadedNo AI model loaded
Technical details & recovery

Transformers.js 4.3.2, onnx-community/ormbg-ONNX, q8, WASM worker. Same-origin Cache Storage model files are reused only when compatible; other runtimes may use inaccessible caches. Never clear all site storage without backing up work.

OPTIONAL CACHE CLEANUP

Delete recognized AI model downloads?

Removed models may need downloading again in other PixelOkay tools. Loaded sessions in other tabs can be affected. Saved projects and unrelated site storage are not targeted; inaccessible model caches may remain.

◎

Local from start to finish

Your uploaded photos are decoded and edited in your browser. AI model files, when requested, download separately, without uploading your product image.

✧

Made for real products

Build consistent marketplace images, social posts and product concepts using a reusable library of editable environments.

▧

Manual control matters

AI segmentation can fail on transparent glass or fine objects. Restore and erase brushes let you correct the mask before export.

GOOD TO KNOW

Frequently asked questions

Does this create entirely new photorealistic scenes with diffusion AI?

No. This edition creates original procedural studio and lifestyle illustrations, or composites your product with your own background photo. The optional AI model is used for foreground extraction, not text-to-image diffusion. We deliberately avoid promising expensive GPU-generation capabilities that are not implemented.

Will the AI model require a download?

Yes. Open AI Model Setup and explicitly select Download & Load Model to permit the optional Transformers.js runtime and model downloads. Download size and runtime depend on the external provider and device. Simply opening the editor or choosing AI Cutout does not download model files; if it fails, you can still use solid-background removal or manual editing.

Is the AI dependency free for commercial sites?

AI Cutout uses Apache-2.0-licensed Transformers.js and the onnx-community/ormbg-ONNX model. Runtime code and model weights load only after the Model Setup download action and then run on your device; there is no paid API. Real-device performance and hosted model delivery still need verification.

Are my product photos uploaded to a server?

No. The app uses local browser Canvas APIs and an optional on-device inference package. External requests are for the model/runtime code and weights only. If you choose a custom background, that image remains local too. The product is never sent to a remote image-generation API.

What inputs and limits are supported?

PNG, JPEG and WebP images that your browser decodes, up to 12 million decoded pixels and 20 MB per file. PNGs with transparency are ready for composition. You can export PNG or JPEG at 800, 1200 or 2000 pixels wide. Large inputs may use more device memory.

Using AI Product Scene Generator

Import a product image, choose optional AI or a supported local cutout method, then refine the mask. Select a procedural scene or your own background photo and adjust the composition. Export images or settings; the settings JSON does not contain photos or masks.

Canvas presets are procedural illustrations, not diffusion-generated photographic scenes. The tool does not generate camera views, 3D geometry or physical relighting. Optional cutout requires explicit model setup with Transformers.js/ORMBG assets; results vary, and cancellation terminates the app-owned inference worker and discards stale results, but cannot guarantee that external CDN/network transfers have stopped immediately. Manual refinements can be necessary. File/output budgets and browser memory apply.

References: ORMBG model · Transformers.js runtime

Workflow and output choices for AI Product Scene Generator

Import a product image, choose a manual or supported cutout method, refine the mask and compose it over a procedural scene or your own background photo.

Image exports contain the composition. Settings JSON retains configuration but not the photos or masks; keep source images separately.

Example workflow

  1. Import a product photo and create a cutout with a supported manual or local method.
  2. Optionally configure ORMBG to suggest a mask, then inspect and refine its edges.
  3. Choose a procedural scene or custom background photo, adjust placement and export the composite.

Authored example: Arrange a fictional soap package over one of the procedural studio presets, then compare it with a user-selected photo background. The presets are illustrations, not generated product photographs.

Limits and model handling: The optional ORMBG cutout loads Transformers.js and model assets after setup; results vary and may need manual refinement. Procedural scenes are not diffusion-generated photographs, 3D geometry, new camera views or physical relighting. Cancellation discards stale app results but may not immediately stop external transfers.

AI Product Scene Generator questions

Does the tool generate a new photographic background from a text prompt?

No. It offers procedural scene presets or a background photo you provide; it is not a diffusion image generator.

Does the settings JSON include the product image or mask?

No. It stores settings only. Keep the source image and any needed mask or exported composite separately.

Related tools