Product photo skill · version 1.0.0
White backgrounds.
Keep your product.
Make a white background without asking a generative model to repaint your logo. AI supplies the alpha mask; the default raw composite uses your color-managed source photograph.
Start here
Free Lite + notebook
$0
- Process one photo at original resolution.
- Automatic mask, ICC color management and pixel checks.
- Local CPU or compatible NVIDIA GPU.
- Optional free notebook in your own cloud account.
Download the Lite ZIP first, then follow the README or upload the ZIP and your photo in Colab. No payment required. Free cloud capacity is not guaranteed.
For repeated work
Full package
$9.99 USD · one time
- Everything in Lite.
- Batch manifests and canvas presets.
- Optional programmatic contact shadows.
- Optional tone lift with measured color checks.
Full ZIP delivery through the Whop Files tab after checkout. Download access: 120 days from purchase. Keep your own copy. No continuing updates promised.
Choose the tools you need
| Feature | Lite · Free | Full · $9.99 |
|---|---|---|
| Single photo + automatic mask | Yes | Yes |
| ICC → sRGB + pixel checks | Yes | Yes |
| Batch manifests | — | Yes |
| Canvas presets | — | Yes |
| Contact shadow + tone lift | — | Optional |
| Buyer-owned cloud notebook | Yes | Yes |
From download to one checked image
- Install. Unzip the kit, create a fresh virtual environment with 64-bit Python 3.11 or 3.12, then run
python scripts/install.py. It selects CPU or NVIDIA dependencies. No model weights are included. - Process. Run
python scripts/process.py --image product.jpg --out work/new-photo. A fresh output directory is required. The first processing run downloads and checks the pinned MIT-declared BiRefNet-Lite ONNX model. - Inspect. Require a successful exit and
passed: truein the reports. Compare the original against the output at 4×, especially text, logos, holes and edges.
What “keep your logo” means
Default raw processing preserves the opaque product core relative to the EXIF-oriented, ICC-converted sRGB reference. It does not use model-generated product RGB. Soft edges are composited according to alpha; a wrong mask can still remove real detail. Optional Full tone lift changes pixels and has separate color gates. Inspect every output.
Canvas presets expand the canvas without shrinking the product pixels; preset names do not certify marketplace compliance. Contact shadows are programmatic and optional. They require a human check for realistic contact and perspective.
Hardware, privacy and test boundaries
Local CPU is standard; a compatible NVIDIA system automatically chooses CUDA. GPU setup errors are explicit. Local images stay local; only the approximately 224 MB model is downloaded and cached. The optional free Colab notebook uploads chosen inputs to your own cloud account. Use private inputs; quotas and free GPU availability are not guaranteed.
In eight Linux CPU/CUDA runs on two 4032 × 3024 photos, CPU commands took 23.82–26.37 seconds/photo and used 6.57–6.72 GiB peak process-tree RAM. This is not a buyer timing guarantee. Plan to start testing with 16 GB RAM, not a validated minimum; 8 GB may be insufficient. Windows clean installation and interactive Colab upload/download remain unverified. The ZIP includes synthetic examples, instructions and scripts; no private photos, model weights or paid service credits.
License, delivery and refunds
You may use and modify the kit for your own and client product-photo work, including commercial work. You keep rights to your images. Do not resell or redistribute the Full kit. Unmodified Lite may be shared without charge with notices intact. Third-party components retain their own licenses. Read the complete kit license.
Full costs $9.99 USD once, delivered as a ZIP through Whop Files. Download access lasts 120 days from purchase; keep a backup of your downloaded copy. No continuing updates or ongoing support are promised. For product/access or refund requests contact support@miloagents.shop; payment errors go to Whop. Whop Buyer Terms and mandatory consumer rights apply. No separate money-back promise is added.
Watch the story and get the EP4 workflow · Free kit repository