hjLabs AI Playground
Starting the on-device runtime…
CLIP embeds images and text into the same space, so you can score a photo against any list of descriptions you write — "a product photo on a white background", "a blurry screenshot", "a person wearing glasses" — with no training data at all. SigLIP is the stronger successor and TinyCLIP the small fast option. This is the practical tool for tagging, filtering and de-duplicating an image library when your categories are specific to you and exist in no public dataset.
zero shot image classificationclip demo onlinesiglip browsercustom image labels aiimage tagging without trainingtinyclip democlip similarity tool
A lot. CLIP was trained on captions, so "a photo of a golden retriever" usually scores more reliably than the bare word "dog". Keep every candidate label in a consistent phrasing.
Yes — embed each image once, keep the vectors, then compare against a text query. That is exactly how semantic image search is built, and every step of it can run here.
No. The model weights are downloaded from Hugging Face to your browser once, and every inference after that runs on your own GPU through WebGPU. Your text, images and audio are never sent anywhere — hjLabs.in has no inference server and no way to see your input.
Small models do. Recent mobile Chrome and Safari expose WebGPU, but memory is the binding limit — stay under roughly 200 MB of weights on Android, and expect iOS to reclaim the tab sooner. The catalog shows every download size up front so you can pick something that fits.
That depends on the individual model, not on this site. Each model page links to its Hugging Face repository, where the licence is stated — most of the catalog is Apache-2.0 or MIT, a few are research-only. We add nothing on top: hjLabs.in claims no rights over anything you produce here.