Short, specific walkthroughs. Each one says what you will have at the end, what to click to get there, and where the tool actually fails — which is usually the part that saves you the most time.
You need an account to run a job — an anonymous generate button is a free compute faucet one forum link away from being someone else's render farm — but every word of these guides is readable without one.
A captioned image, at the size you supplied, with text that says exactly what you typed.
A PNG with a real alpha channel, or your subject standing in a scene it was never photographed in.
An old scan, a compressed download or a video still with its facial detail rebuilt.
The same picture at four times each dimension, with edges and texture reconstructed rather than interpolated.
A short MP4 (or GIF) generated from your description, or from a photo you supply as the first frame.
Your photo back with exactly the change you asked for, and nothing else moved.
A conversation with a model that runs on our infrastructure, and a clear idea of which parts of its answer you have to check.
Four limits, each one traceable to a law rather than to our taste. Everything outside them is a setting on your account, and politics is not on the list.
What the free tier actually includes, how long results are kept, where the AI-generated label appears, and where to cancel.
The same stack over an OpenAI-compatible API and Comfy, so an editor plugin or a script works without anything from us.
Be furious. Be ferocious. Express yourself boldly.