Frequently Asked Questions

Yes. The core workflows run locally in your browser, so your file contents do not get uploaded to our servers for processing. After the required model is downloaded, many workflows can continue offline.

WebGPU is a modern browser graphics and compute API that can use your device's GPU for on-device AI tasks. PrivaKit is built around browser-side inference, and some workflows may use WebGPU where available while others fall back to stable local runtimes like WASM.

WebGPU support now depends on a mix of browser version, operating system, GPU, and hardware acceleration settings. Chromium browsers generally offer the broadest support, Firefox has shipped support on some Windows configurations, and Safari 26 adds support on newer Apple platforms. We recommend using the latest browser version and checking whether WebGPU is enabled on your device.

The first run downloads the model required for that workflow, typically around 30MB to 100MB depending on the task. The browser caches those files locally so later runs are much faster and can work offline.

Processing speed depends on your device, browser, file size, and which local runtime the workflow uses. Faster GPUs help, but some workflows currently prefer stable browser-side runtimes such as WASM over WebGPU.

Not yet. The current tools focus on one-file-at-a-time local processing. If batch workflows become available, we will document them explicitly in the product UI and FAQ.

Yes, PrivaKit is an open source project. The AI models we use are all from the open source community, including RMBG-1.4 (background removal), Real-ESRGAN (upscaling), and Whisper (transcription).

Mobile support is still mixed. Browser support, hardware acceleration, memory limits, and thermal constraints vary a lot across devices. For the most reliable experience today, we recommend using a desktop or laptop browser.