Subtitles QA Engine: Intelligent Automated QC Tool for Multilingual subtitles

Context

CaixaForum+ receives subtitles in various languages from external providers and needs to ensure their quality before publication. Currently, the review and quality control process is manual and time-consuming, as it requires validating spelling, formatting, duration, synchronisation, and accessibility standards.

In addition to conventional subtitles, the platform includes subtitles for the deaf and hard of hearing (SDH), which require specific quality criteria.

Challenge description

Develop a technological tool capable of performing a complete automated quality check of subtitle files (.srt, .vtt and similar formats), applicable to both standard subtitles and accessibility content (SDH), integrated into the CaixaForum+ publication workflow.

Objectives:

• To detect linguistic and technical errors in multiple languages.

• To validate professional subtitling standards (characters per second, number of lines, line breaks duration, size, etc.).

• To verify synchronisation between audio and timecodes.

• To detect overlaps, incorrect formats, impossible intervals, or invalid timestamps.

• To identify frames with embedded text or graphics that interfere with subtitle placement.

• To evaluate and validate SDH according to international accessibility standards.

• To generate automated reports and enable auto-fix suggestions.

Solution requirements

The proposal must include:

1. A functional MVP or demo capable of analysing and reporting errors in at least 2–3 languages.

2. Detection of spelling, technical, semantic, formatting, and accessibility errors.

3. Compatibility with SDH subtitles.

4. Automated reporting system with error classification by severity.

5. Scalability and API integrationwith editorial workflows or CMS.

Expected impact

• Reduction of manual review time by up to 70%, evolving from manual to automated quality control.

• Guarantee of quality and consistency across