hjLabs AI Playground
Starting the on-device runtime…
NLLB-200-distilled covers 200 languages in a single model; the Opus-MT family gives smaller, faster models for one specific language pair. Both run here on your GPU, which means no character cap, no per-word pricing and no document leaving your machine — the usual reason people cannot paste a contract or a medical note into an online translator. Quality sits below the big commercial engines on rare pairs and is competitive on common ones.
offline translatorfree translation no limitnllb-200 demoopus-mt browserprivate document translationneural machine translation onlinetranslate without apilocal ai translator
Opus-MT is usually better and much smaller for a single well-supported pair such as English↔German. NLLB-200 is the one to use for low-resource languages or when you need many pairs from one download.
Yes, once a model is cached. Weights are stored in the Cache Storage API under the key "transformers-cache", so a reload with no network still runs the model. Call "Make storage persistent" first — browsers may evict non-persistent caches under disk pressure.
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.