BackgroundErase licensing is designed for teams that want more control than a standard hosted API plan. In most cases, the choice comes down to two main paths: self-hosted and on-device. Both let you run the model outside of the public BackgroundErase API, but they solve different problems and fit different product strategies.
The easiest way to think about it is this: self-hosted is for teams that want to run the model on their own infrastructure, while on-device is for teams that want to ship the model directly inside a customer-facing application such as an iPhone or Android app.
Simple summary: self-hosted is usually the best fit for server-side control and flexible deployment, while on-device is usually the best fit for native app experiences, lower runtime latency, and processing directly on the user’s device.
The two main licensing options
BackgroundErase currently centers enterprise licensing around these two deployment models:
- Self-hosted: your team runs the model on your own hardware or cloud environment. That could mean a local server, your own cloud provider, a dedicated GPU box, or even a local machine like a MacBook depending on your use case.
- On-device: your team ships the model directly inside a mobile app. We offer ready-to-ship Core ML and TensorFlow Lite packages for this path.
Both options can be licensed in ways that fit long-term commercial deployment rather than pay-per-call hosted usage.
Self-hosted at a glance
Self-hosted licensing is for customers who want direct operational control. In this model, the BackgroundErase model runs entirely inside your own environment. That could be an on-prem server, an EC2 instance, a private cloud deployment, a Kubernetes cluster, a workstation, or a local machine used for a specific internal workflow.
This option is usually attractive for companies that already have server-side infrastructure, want tighter control over scaling and routing, or need to keep the runtime inside their own environment for technical, business, or privacy reasons.
- Runs on your own hardware or cloud
- Fits backend and batch-processing workflows
- Works well for server-side product integrations
- Gives your team more deployment and infrastructure control
- Available as yearly or perpetual licensing
This is often the right choice for internal media systems, SaaS backends, marketplaces, ecommerce operations, and private inference environments.
On-device at a glance
On-device licensing is for teams that want the model to live directly inside the end-user application. Instead of sending each image to a server for inference, the app performs the work locally on the device itself.
We currently offer Core ML and TensorFlow Lite packages that are ready to ship into production mobile apps. This path is usually more expensive than self-hosted licensing because it creates a different kind of product value: lower latency, local execution, and a more native app experience.
- Designed for iOS and Android app deployment
- Ready-to-ship Core ML and TensorFlow Lite packages
- Lower runtime latency for end users
- Excellent fit for native mobile UX
- Generally priced above the self-hosted option
Yearly versus perpetual licensing
The licensing structure can usually be split into yearly and perpetual options.
A yearly license is usually the better fit for customers who want lower upfront commitment, easier budgeting by contract cycle, and a more flexible path as their deployment evolves. A perpetual license is usually the better fit for customers who want to make a one-time investment in long-term usage rights for a given deployment path.
Both self-hosted and on-device conversations can involve yearly or perpetual structures depending on the customer’s business model, rollout timeline, and procurement preference.
Optional maintenance packages
Yearly and perpetual licenses can also come with an optional maintenance package. This is generally the right choice for customers who want enterprise support, upgrade guidance, implementation help, and a more active long-term relationship instead of a simple one-time delivery.
A maintenance package can make sense when the model is part of a serious production workflow and the customer wants help with deployment questions, future updates, rollout support, or ongoing technical coordination.
- Enterprise support access
- Implementation and rollout guidance
- Ongoing technical relationship
- Help with upgrades and future model revisions
- More predictable long-term support coverage
Which option is right for you?
In most cases, the decision comes down to where you want inference to run and what kind of product experience you want to build.
- Choose self-hosted if you want to run the model on your own server, cloud instance, or internal machine.
- Choose on-device if you want to embed the model directly inside a mobile application and prioritize local execution.
- Choose yearly licensing if you want a lower-commitment recurring contract model.
- Choose perpetual licensing if you want a longer-term ownership-style licensing path.
- Add maintenance if you want enterprise support and a stronger ongoing working relationship.
Common buyer profiles
These are the kinds of teams that usually map to each option:
- Self-hosted buyers: SaaS platforms, enterprise backends, catalog operations, media pipelines, private internal systems, and companies that want the model running on infrastructure they control.
- On-device buyers: mobile apps, creator tools, camera apps, ecommerce seller tools, and consumer products where the feature needs to feel instant and native on the device.
Related licensing articles
This overview is the starting point. From here, customers usually go deeper into the specific path they are evaluating, such as self-hosted deployment details, on-device packaging, yearly versus perpetual licensing, or support and maintenance.
The simplest version
BackgroundErase licensing mainly comes in two forms: self-hosted for customers who want to run the model on their own infrastructure, and on-device for customers who want to ship ready-to-deploy Core ML or TensorFlow Lite packages inside mobile apps. Both can be licensed yearly or perpetually, with optional maintenance for enterprise support.
Contact sales
If your team is deciding between self-hosted and on-device licensing, the best next step is to visit backgrounderase.com/enterprise and tell us how you plan to deploy the model.