Evaluation of AI-Generated Translated Captions

This technical report presents an independent evaluation of AI-generated translated captions across five platforms: Google Meet, Microsoft Teams, Zoom, DeepL Voice for Microsoft Teams, and DeepL Voice for Zoom. The study aims to assess translation quality and caption stability, two critical dimensions for user experience in real-time multilingual meetings. The evaluation involved 28 professional linguists who conducted blind assessments of translated captions across 14 language combinations. The results indicate that DeepL Voice products achieved the highest quality scores, significantly reducing critical translation errors compared to other platforms. Specifically, DeepL Voice for Zoom scored 96.4 out of 100, while DeepL Voice for Teams scored 96.3. The report also details the methodology used for capturing and analyzing captions, including automated measurements of caption stability. The findings highlight the performance differences across platforms, emphasizing the importance of both translation accuracy and visual stability for effective communication in global meetings.