Pro Model Self-Host
- Edge refinement: hair and fine detail refined after the first pass (MAE 0.0276 on our 60-image benchmark)
- Refine: send Keep / Remove masks with an image to fix a cutout
- Runs on CPU: a 2.1 GB Docker image, no GPU needed
- Images stay with you: nothing is sent to withoutBG
- €3,600 + VAT, one-off flat fee: perpetual license with 12 months of updates (license terms)
Self-hosted · CPU · Commercial license
Photo by Hasan Almasi on Unsplash
How it compares#
The three withoutBG models scored on the same 60 photos with ground-truth alpha mattes. Lower is better on every metric. Pro Model Self-Host's MAE is 5.6% lower than the free open-weights model and 15.0% higher than the Pro Model API.
| Model | MAE | Edge (MGE) | Runs on | License |
|---|---|---|---|---|
| Pro Model Self-Host | 0.0276 | 0.0369 | Your CPU servers | Commercial license, one-off fee |
| Pro Model API | 0.0240 | 0.0274 | withoutBG cloud (Frankfurt) | Pay per image |
| Open weights | 0.0292 | 0.0452 | Your infrastructure | Free, open model license |
Examples#
Sample outputs from Pro Model Self-Host. Switch between transparency, chroma key, and alpha matte on each result.
Photo by Niko Tsviliov on Unsplash
API or Self-Host?#
Two models, built for different places to run.
- Pro Model API: runs on GPUs we operate, at full resolution. Nothing to deploy; you pay per image.
- Pro Model Self-Host: a model built to run on CPU, with the same edge refinement and Refine. You deploy and scale it on your own servers, pay €3,600 + VAT once, and images never leave your network.
When to choose Pro Model Self-Host#
- Images can't leave your network. Contracts, regulation or internal policy rule out a third-party API. Pro Model Self-Host runs inside your own infrastructure.
- You run CPU servers. It runs on the servers you already have, so there is no GPU server to rent.
- You need more than the open-weights model. The open-weights model is free to self-host; Self-Host adds edge refinement and Refine on top.
If you'd rather not run infrastructure, Pro Model API is fully managed.
Requirements#
What a server needs to run Pro Model Self-Host. It runs on ONNX Runtime on the CPU.
| Delivery | A Docker image (2.1 GB) with the model inside. No download at start. |
|---|---|
| CPU | x86_64. It uses every core by default; set WITHOUTBG_ORT_THREADS to cap it on a shared server. 4 cores minimum, 8 or more recommended. |
| GPU | Not needed. |
| RAM | 8 GB minimum, 16 GB recommended (a container peaks at about 5.5 GB). |
| Disk | 5 GB free for the Docker image and the runtime. |
| Software | Linux (x86_64) with Docker. |
| Network | None to withoutBG. It runs offline and sends no telemetry. |
| Images | Any size. The matte is estimated at up to 1024 px on the long side and resized back to the original size; a setting raises that to 4096 px for finer edges on large photos, at more CPU time. |
| Scaling | Each container processes one image at a time and queues the rest. Run more containers or servers for more throughput. |
Get a license#
The license is €3,600 + VAT, one-off. Tell us about your deployment target and your timeline, and we'll reply with the Order Form. The license terms page sums up what the license allows.