Testing and publishing methodology

This page explains the evidence Aiviko Team uses before describing a tool as supported, local, or ready for general use. It is a practical product-testing process, not a certification or a promise that every input will produce the same result.

What we check

For each material release we run automated type, lint, unit, production-build, and route checks. We also exercise the main workflow with representative supported inputs in current desktop and mobile browsers. We check keyboard access, visible focus, narrow screens, empty states, errors, downloads, and navigation to help and policy pages.

Each tool has different test inputs. QR output is scanned after generation. Conversion is checked with the formats named in the interface and with boundary cases such as large dimensions or page ranges. OCR is checked with clear Korean and English samples, mixed numbers, rotated or low-contrast text, and intentionally unsupported scripts. Speech, subtitle, and background-removal results are reviewed with short representative media; model output is never described as guaranteed.

Local processing and network claims

“Processed in your browser” means the selected tool input is handled by client-side code and is not intentionally sent to an Aiviko processing endpoint. It does not mean the page is completely offline. Application chunks, fonts, WASM runtimes, speech or vision models, analytics, advertising, and infrastructure services may still be requested from Aiviko or the providers listed in the privacy and cookie policies.

We review network behavior during development and keep advertising disabled until the production consent setup is marked ready. A network audit must be repeated after changes to hosting, CDN, analytics, ads, consent tooling, models, or third-party scripts.

Accuracy, speed, and limits

Results depend on the input and device. OCR accuracy changes with script, resolution, lighting, layout, and model support. Speech results change with language, noise, speakers, compression, and model choice. Large files can exceed browser memory even when they fit the visible file-size limit. Performance claims are therefore descriptive, not guaranteed, and we document reproducible limits instead of publishing a universal success rate.

Editorial review and corrections

Help content is maintained by Aiviko Team and reviewed when the matching tool changes. We prefer first-hand tests, implementation details, and upstream project documentation. Generated drafts are not treated as evidence on their own. If instructions are inaccurate or an attribution is missing, send the page URL and reproduction details to support@aiviko.app. Confirmed corrections are applied to the page or recorded in the changelog.

Last methodology review: August 27, 2026.