AI tools that run in your browser
Each tool downloads a small neural network once and runs it on your own device: on the graphics card through WebGPU, or on the processor if WebGPU is not available. Your photos, recordings and texts stay with you.
The file is processed on your device and is not uploaded to a server.
- Remove backgroundTransparent PNG from any photo, made on your device
- 3D photo & depth mapDepth map and a moving parallax photo from one picture
- Select objectClick an object to cut it out with Segment Anything
- Speech to textWhisper transcription with TXT and SRT, without upload
- Image to text (OCR)Read text from screenshots, photos and scans
- Private translatorTranslate text on your device, no server involved
- Local AI chatA small language model that answers offline
- Semantic searchFind lines in a list by meaning, not by exact words
Questions
- Are my files really not uploaded?
- Yes. The page downloads the model from Hugging Face and the processing library from jsDelivr. Your file is read by the browser and processed in it; nothing is sent to our servers. You can check this in the Network tab of the developer tools.
- Why do I need to download a model?
- The neural network is a file of tens to hundreds of megabytes. It is downloaded once when you press the button and then stays in the browser cache, so the next time the tool starts in seconds.
- Which browser works best?
- A recent Chrome or Edge on a laptop: they support WebGPU and run models on the graphics card. Firefox and Safari also work, using the processor, which is slower.