Plain-English explanation
Hosted generation sends a prompt or message to provider infrastructure and returns the generated result. You never download anything larger than the app itself; the actual model, often far bigger than anything a home GPU could run, lives entirely on the provider's infrastructure.
This is the default approach for the large majority of commercial AI companion apps, chatbot platforms, and image or video generators, largely because it's dramatically easier for a typical user to access than setting up local generation themselves.
How it works
A request, a prompt, an uploaded reference image, a chat message, travels from your device over the internet to the provider's servers, usually through an API whether or not you interact with it directly. The server runs inference using the provider's own hardware and sends the output back. Usage is typically metered, a subscription tier or credits spent per generation, since the provider is paying for the compute behind every request, not you.
Adult-AI use
Hosted adult AI tools run generation on provider infrastructure. It offers convenience and access to large, capable models without any hardware investment, but every prompt and generated output typically passes through and may be logged, reviewed, or retained by the provider under its own privacy policy, a real consideration for adult content specifically. Hosted platforms also enforce their own content policy and NSFW filter regardless of what a user wants to generate, and accounts can be restricted or banned for violating it; age verification requirements are also typically enforced at this layer, since the provider bears legal responsibility for what its service produces. None of this makes hosted generation unsafe to use, reputable providers publish policies specifically to define these limits, but it does mean the provider, not the user, holds final control over what's possible and what happens to submitted data.
How it differs from local generation
Hosted generation trades control and on-device privacy for convenience, larger model access, and no hardware requirement. Local generation trades that convenience for privacy and freedom from a platform's content policy, at the cost of hardware investment and setup effort. See local generation for the other side of this comparison.
Common uses
- Most consumer AI companion and character chat apps.
- Most consumer text-to-image and text-to-video web tools.
- API-based access to a provider's model for developers building their own front end.
Limitations and misconceptions
- “Hosted” doesn't automatically mean lower quality than local: many hosted platforms run models far larger than consumer hardware could ever handle, often the newest and most capable versions available.
- Deleting content from your own device doesn't delete it from a hosted provider's servers: retention is governed by that provider's own policy, not by a local action.
- Free hosted tiers commonly come with tighter content restrictions, lower generation limits, or lower output quality than paid tiers on the same platform.