Welsh translation and lip-sync for video production.
The Llais project represents a pioneering initiative aimed at exploring the potential impacts and ethical implications of utilising artificial intelligence (AI) for lip-syncing in Welsh to English video content.
This project seeks to assess how AI technologies can be fine-tuned and packaged for various markets, thereby enhancing the accessibility and distribution of Welsh-language media globally.
Objectives and Scope
The project is focused on the following objectives:
+ Investigating the effectiveness, ethics, and cost-efficiency of AI-driven lip-syncing in Welsh-language video production.
+ Providing opportunities for translators and dubbing artists to create new services tailored for training, corporate video markets, and television production companies.
+ Engaging with industry stakeholders to understand concerns and insights related to AI in the creative sector.
Summary
The Llais project is positioned to drive significant advancements in the Welsh-language media landscape by harnessing AI technology responsibly and ethically.
By addressing industry concerns and incorporating stakeholder feedback, Llais aims to enhance the accessibility and quality of Welsh content while promoting inclusivity and representation on a global scale.
A comprehensive report detailing testing outcomes and lessons learned will be provided at the end of the initiative.
Actor: Tom Mumford; Camera: Ian Smith, Auntie Margaret productions; Tech partner: Robin Moore, Shwsh; Project mgt: Aled Jones, Y Pod Cyf
Click here to download the full Llais report
Actors: Melanie Owen, Tom Mumford, Camera: Ian Smith, Auntie Margaret productions.
Video examples:
The videos below are for researching the effectiveness of AI in synchronizing Welsh and English audio and were recorded for the Llais research and innovation project.
Testing Process
For the testing phase, Llais filmed two actors: a female and a bearded male, both proficient in Welsh. The filming took place in an innovation studio in Cardiff, with a professional camera operator overseeing the process.
The actors were recorded in both languages to facilitate effective comparison and assessment. Four different AI models were employed during this testing phase to evaluate their effectiveness and performance.
A comprehensive testing report detailing the findings and lessons learned from this phase can be found in Appendix A of the Llais Final Report.
The Llais project has yielded several important insights from both the research conducted and the testing process. These learnings will play a crucial role in shaping the project’s development and informing the creation of a viable business model.
Which is AI generated? News Comparison Reveal.
Actor A Walking: AI Model 3 – Filmed in English dubbed in Welsh.
Actor B Walking: AI Model 3 – Filmed in Welsh, dub in English
News items: Actor B – AI Model 3: Filmed in English, dub in Welsh
Conversation: AI Model 3 – Filmed in Welsh, dub in English
AI Model Comparison: Walking
AI Model Comparison – News: 4 AI Models
Model 4 doesn’t recognise Welsh.
AI Model Comparison: Facial Hair
Project Summary
The Llais project’s research illustrates the significant potential for AI-driven lip-syncing in expanding the reach and impact of Welsh-language media.
With ethical considerations, accessibility improvements, and a market-ready product, Llais could reshape how Welsh content is experienced locally and internationally, making it more inclusive and widely accessible across different media sectors.
The Llais project has garnered interest due to its potential positive impact, although there remains an awareness of its implications for the creative sector.
There is a consensus that Welsh language initiatives should lead in the adoption of technological advancements within the creative industries, and individuals within the sector intends to monitor the development of the Llais project closely.
Credits:
Project mgt: Aled Jones, Y Pod Cyf
Tech partner: Robin Moore, Shwsh
Actor: Tom Mumford
Actor: Melanie Owen
Film: Ian Smith, Auntie Margaret productions