The Interface Members Actually Touch
Short answer: The model is not what a person touches. They touch the interface: chat with streaming, history and a login, and one simple screen where chat, images and video live. Building that screen — and keeping it working — is its own project, separate from renting the GPU or picking the model.
What the interface has to do
Chat needs a front end with streaming, history and a login. Streaming, because an answer that arrives all at once after a long pause feels broken even when it is not. History, because a conversation you cannot return to is not an assistant. A login, because the build holds your email, calendar and data behind it.
The Freeborn homepage describes the finished result as top open models for pictures and video in one simple screen. One screen is the standard: the person using it should not have to know which tool is running underneath.
The ready-made starting point
Open WebUI is a ready-made chat front end, named on the pillar page. Image and video run in ComfyUI, which you drive through its API — so the interface work includes making a tool built for operators feel like one simple screen for a member.
A ready-made front end shortens the start. It does not remove the project: the interface still has to be connected to the harness, given a login, and kept working when the pieces underneath it update.
Why it is its own project
The pillar page puts the interface and the harness together as the largest step of the build, and the interface is the half a member can see. When an update to vLLM or ComfyUI breaks everything on a Tuesday, the interface is where the breakage shows up first.
That is also why the guide's honest order is chat before images, and images before the agent: each layer gives the interface one more thing to present simply, and each one is easier to make simple when the layer before it already works.
The hours, as estimated on the pillar page
The interface and harness together: about 40 to 80 hours the first time, then about 2 to 4 hours a week. These are the pillar page's estimates, not measured data, and they assume you have built software before. If you have not, add time.
Related DIY pages
Back to the pillar: What it takes to do this yourself
Adults only (18+). What it takes to do this yourself · Apply