Automating multi-language subtitles

The challenge

A large library of English-language educational videos required localisation into French, German, and Dutch for a corporate learning platform.

Each video contained an English VTT subtitle file that needed translation while preserving:

• exact timestamps
• formal language
• brand names
• speaker names
• domain-specific terminology

The challenge was maintaining consistency and quality across the subtitle files covering topics such as leadership, AI, and wellbeing – without introducing translation errors or requiring extensive manual review.

The solution

To optimise the localisation process, I designed a structured AI-powered subtitle translation workflow using Claude AI.

A dedicated Claude Project was configured with translation standards, terminology rules, quality requirements, and formatting instructions. This allowed every translation session to start with the same institutional knowledge and quality framework.

The workflow automatically generated:

• French subtitles
• German subtitles
• Dutch subtitles

from a single English VTT file while preserving timestamps and formatting requirements.

The solution transformed a repetitive localisation process into a scalable and repeatable workflow.

Workflow overview

English VTT file
        ↓
Claude project
        ↓
Translation rules
        ↓
Timestamp validation
        ↓
FR / DE / NL outputs
        ↓
Ready for publishing

The workflow embedded terminology handling, formatting requirements, timestamp preservation, and quality controls directly into the process, reducing the need for manual QA.

Results and impact

Localisation efficiency

• three languages generated from a single workflow
• no repeated instructions required after initial setup
• outputs ready for immediate publishing
• scalable across multiple content domains

Quality improvements

• 100% timestamp preservation
• consistent terminology handling
• formal language standards maintained
• reduced need for post-translation review

Workflow improvements

• eliminated repetitive prompting
• reduced context-switching between projects
• standardised localisation processes
• enabled scalable multilingual content production

The largest efficiency gain came from eliminating manual QA overhead, and repeated instruction management across translation projects.

	Traditional workflow	AI-Powered workflow
Languages per pass	1	3
Translation instructions	Repeated manually	Stored permanently
Timestamp validation	Manual checking	Built into workflow
Quality control	Post-review	Embedded rules
Output readiness	Additional formatting required	Ready for publishing
Scalability	Moderate	High