Arthub.ai

Each upload shows the artwork together with the text that created it. Open a post and you'll see the full prompt, and often the generation settings like sampler, steps, seed, and model. That makes the feed a working reference, not just a gallery. If a look catches your eye, the recipe is right there. That's the whole idea.

Arthub runs its own text-to-image generator, reachable from the Generate link in the top navigation. You type a prompt and it produces an image without leaving the site. It's a convenience feature rather than a full studio, so expect a simpler set of controls than a dedicated desktop app would give you, and don't plan on fine-tuning every setting before you hit generate. For a quick draft, that's usually enough.

The feed is split into categories such as Fantasy, Characters, Low Poly, Sci Fi, and Furniture, with a separate search that indexes prompt text. Searching a phrase like "cinematic lighting" surfaces every post that used it, which means you can reverse-engineer a style from a single phrase you half-remember. This is the fastest way to find examples of a specific style or subject. Type it, hit enter, done.

Beyond individual posts, Arthub keeps a Prompt Library that strips the images away and lists the prompts themselves. It's built for people who want to read and collect wording, not scroll pictures. Copy a few entries and you have a starting template for your own work, which beats opening a blank generator and hoping for the best. No images required.

Every user gets a profile page that collects their uploads. You can browse a single creator's entire body of work, from one experimental prompt to a long series. For anyone building an audience, the profile acts as a portfolio tied directly to the method behind the art. Simple, and it works.

Under many posts you'll find the generation parameters: model name, seed, resolution, sampler, CFG scale, and even LoRA hashes. That level of detail lets you reproduce a result far more closely than the image alone would allow, since you can match the seed and sampler and land on nearly the same output. It's the kind of transparency most galleries skip, and it's the feature that turns a pretty picture into something you can actually learn from.