QACAT - Catch translation issues before your users do
QACAT is a hybrid translation QA platform combining automated rule checks, AI analysis, and expert human review. Upload screenshots and review translations in real product UI — built-in OCR pulls the text for you. Every run gives a structured, scored report with severity breakdowns and an AI summary of what to fix. Works across 100+ languages. Powered by Alconost.
Replies
finally a QA tool that treats quality as something you track over time rather than a one-off score per project!
How does the OCR handle languages with non-Latin scripts or right-to-left text like Arabic or Hebrew?
Alconost Localization Lab
@margarita_tsygankova Thanks for asking. Good question indeed! We use two OCR engines. One supports Arabic, Hebrew (AI assisted), and other non-Latin scripts well (no AI involved, for NDA safe mode), while the other is optimized for Latin-based languages. We select the most suitable OCR approach based on the content being processed.
Hi! I am wondering how this is different from just running a quick LLM QA check myself?
Alconost Localization Lab
@olga_sapach Quick LLM QA checks are absolutely a valid modern approach, and I use them myself. The difference is that QACAT combines multiple quality layers rather than relying on a single AI opinion.
Depending on the workflow, that can include deterministic QA checks, + LLM-based evaluation, +human validation, and + deeper review workflows such as LQA or LQT.
In our experience, even the best and most expensive LLM models are excellent at finding some types of issues, but they still miss context, make incorrect assumptions, or generate false positives. That’s why we treat AI as one layer of the quality process rather than the final authority.
superlog
Scoring per ai engine is useful. Most teams try multiple ai engines, and knowing which engine performs best per language is valuable data.