Most sellers open Photoroom for the cutout: drop in a photo, lose the background, set the product on white or a colour with a soft shadow, and export at the size the shop wants. Open models such as BiRefNet now do that cutout well on an ordinary laptop, so an agent can wrap a batch tool around one in a single sitting. What stays behind is the generative side (AI scenes, fashion models, video) and the phone app you shoot and edit in. For the scenes you would call an image API and pay per picture.
Before you buildOnly a few photos a week? The free Photoroom account allows up to 100 exports a month with the core tools.
Build me a local product-photo studio to replace Photoroom. Requirements: - Python 3.12 with FastAPI serving one plain HTML page on http://localhost:4700, plus a CLI: `studio batch ./in --preset amazon`. - Cutouts with rembg and the birefnet-general model; fall back to isnet-general-use on machines with under 8 GB of RAM. Say how large the model is before the first download. - Compositing with Pillow: trim to the product, centre it with a padding setting, add a soft shadow made from the blurred alpha mask, and place it on white, a hex colour, transparency, or a background image from ./templates/. - Presets in presets.json, each with size, background, padding and format. Ship amazon (2000x2000, pure white, product fills 85%), square (2048x2048) and portrait (1080x1350). - Batch: accept a folder or many files, run a worker pool sized to the CPU, show progress, and write ./out/<date>-<preset>/ plus a zip of it. - Touch-up: open any result in the page and paint on its mask with restore and erase brushes; saving re-renders that image. - Never change the originals. Save each batch's settings in job.json next to the output so it can be re-run with another preset. - No accounts, no telemetry, no network calls after the model download, local-first. - Out of scope: AI-generated scenes, fashion models, video and a phone app. - README: install, model download size, rough seconds per image on CPU and GPU, and how to add a template.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Photoroom.
Add AI backgrounds to the product-photo studio in this folder. Requirements: - Read OPENAI_API_KEY from .env and use the images edit endpoint with the composited image and a mask that protects the product pixels. - In the page, a text box for the scene, 1-4 variations, and a cost estimate before anything is sent. - Paste the original cutout back on top of every result so the product itself is never redrawn. - Save results next to the batch with the prompt in job.json. - Update the README with what leaves the machine and what it costs per image.
Add a watch folder to the product-photo studio in this folder. Requirements: - Watch a folder set in .env (default ~/Pictures/ToStudio) with watchdog. - Each new JPEG, PNG or HEIC is processed with a default preset from .env, and the result lands in ./out/inbox/; the original moves to ./done/. - Wait until a file stops growing before reading it, so half-synced files are skipped. - Add HEIC support with pillow-heif. - Update the README with how to sync the folder from a phone (iCloud Drive, Syncthing or AirDrop).
Add upscaling to the product-photo studio in this folder. Requirements: - Use the realesrgan-ncnn-vulkan binary, path set in .env, with the realesrgan-x4plus model. - Upscale only when the trimmed product is smaller than the preset needs, and show which images were upscaled in the batch summary. - Add an "upscale: on/off/auto" field to presets.json. - Update the README with where to download the binary for each OS.
Add a Shopify export to the product-photo studio in this folder. Requirements: - Read SHOPIFY_STORE and SHOPIFY_ADMIN_TOKEN from .env. - Match each output file to a product by SKU in the filename (SKU_1.jpg, SKU_2.jpg) and upload it as a product image through the Admin GraphQL API. - Dry-run first: list the matches and ask before uploading. - Never delete existing product images. - Update the README with the token scopes needed.
Resellers and small shops pay because everything happens on the phone they already shoot with: snap, cut out, pick a template, post. Larger sellers pay for batch work and the generative tools, which turn one plain photo into lifestyle scenes or on-model shots without booking a studio.
Estimate · Cutouts run on your own machine at no cost. The build is token usage on your agent plan; the AI-scenes follow-up adds a per-image charge from whichever image API you choose.
Checked 2026-10-13 against Photoroom's docs: 1, 2, 3, 4, 5, 6
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Free | $0 | $0 | Up to 25 exports a week and 100 a month with the core tools. |
| Pro | $12.99 | $7.50 | 1,000 exports and 8,000 AI credits a month; unlimited core tools (background remover, retouch, preset shadows, resize, enhancer), batch editing, brand kit, up to 50 seats. |
| Max | $34.99 | $20.99 | 3,000 exports and 25,000 AI credits a month; adds the video generator, adjustable AI shadows and the Shopify integration. |
| Ultra | $99 | $82.50 | 10,000 exports and 75,000 AI credits a month; 4K images and video. |
| Enterprise | custom | — | Usage-based annual balance, SSO, SLA and dedicated support. |
| API Basic | $20 | — | Background removal at $0.02 an image; the Image Editing API (backgrounds, shadows, relighting) is $0.10 an image. |
A free account still exists, with up to 100 exports a month, but the pricing page now leads with a trial of the paid plans.
Yes. An AI coding agent such as Claude Code, Codex or Cursor can build a usable personal version in one session with the prompt on this page, running on your own machine.
About $12.99 a month on the Pro plan, $155.88 a year (checked 2026-10-13).
AI-generated scenes and relighting; AI fashion models and virtual try-on; Product video generation; The phone app for shooting and editing on the go; Shopify sync, shared brand kits and team spaces. If any of those matter to you, keep paying.
Yes: rembg (open source), BiRefNet (open source), backgroundremover (open source).