Coursera are one of the world’s leading online learning platforms, committed to providing universal access to high-quality education. With more than 100 million learners, 7,000+ courses, and partnerships with 300+ leading universities and companies, Coursera have reshaped how people learn in the digital age.
But as the platform expanded globally, a critical challenge emerged: how to provide truly localised, emotionally resonant learning experiences for students across languages and cultures without compromising authenticity or incurring unsustainable production costs.
Mustafa Furniturewala, CTO at Coursera, oversees Coursera’s engineering team and AI efforts. His work focuses on building generative AI tools (such as AI dubbing, voice cloning, lip‑sync, etc.) and ensuring that technological advancements serve the core mission of creating personalised, accessible, and effective education for everyone, everywhere.
‘Coursera are on a mission to deliver learning experiences globally at scale,’ Mustafa said. ‘And to do that effectively, we have to remove language barriers, personalise learning, and preserve the quality and intent of the original instruction. That’s where generative AI, and specifically HeyGen, has been transformational.’
Delivering immersive, tutor-led experiences
Before using HeyGen, Coursera had successfully deployed AI to scale text and audio translations. But video localisation remained a formidable obstacle.
“We started with text,” Mustafa said. “That was manageable, but when it came to translating video content with the instructor’s tone, emotion, and lips synced to the new language, it just wasn’t feasible. It was too expensive and too complex to do at scale.”
The breakthrough came when Mustafa and his team discovered HeyGen, specifically voice cloning and lip sync. Coursera implemented HeyGen to generate localised versions of instructor‑led videos in French, Spanish, German, and other languages, preserving not just what’s said, but how it’s said.
‘What was most surprising was how natural it felt,’ Mustafa said. ‘We were able to take a video of an instructor speaking English and make it feel as though they were speaking Spanish, whilst maintaining their tone, cadence, and emotional expression.’
That authenticity was non-negotiable for Coursera. The team had always been wary of AI-generated content that felt robotic or disconnected. ‘We’re not just delivering information. We’re creating immersive, instructor-led experiences,’ Mustafa explained. ‘If the emotion gets lost, the learning suffers.’
HeyGen’s ability to deliver quality at scale without sacrificing emotion or instructor presence was what Coursera had been waiting for.
Realising the ‘magic moment’ with HeyGen
Mustafa reflected that what used to require large investments and time now happens with agility. ‘We always thought video translation at scale would require massive investment in production or technology,’ he said. ‘When the team were able to build this in a matter of a few sprints or weeks, we were surprised by how fast we could get to this kind of scale.’
For instructors, seeing their voice and emotion preserved in translated videos was quite powerful. ‘It felt like them,’ Mustafa said. ‘The translated videos still looked and sounded like our teachers, just in another language.’
The technical achievement was matched by the emotional response. ‘The magic moment for us was when we saw a dubbed video with perfect lip sync and preserved emotion,’ Mustafa recalled. ‘That’s when we realised this isn’t just a tool, it’s a platform for scaling learning without losing what makes it human.’
HeyGen didn’t just improve logistics, it unlocked new creative and operational possibilities. Coursera’s engineers and content creators can now say, ‘I made this,’ not just because they created the educational material, but because they translated and localised it themselves, without relying on third-party help.
‘That was quite surprising,’ Mustafa said. ‘The video was in a completely different language, but it still carried the context, the emotion, and the instructional clarity of the original. It felt natural.’
Improving viewing times and completion rates
Since implementing HeyGen, Coursera have achieved measurable gains in learner engagement, platform accessibility, and educational outcomes.
Key results:
- 40% increase in video watch time – In regions such as Latin America, localised video content significantly increased total viewing time.
- 25% boost in course completion rates – Students were more likely to finish courses when video content was delivered in their native language.
- Scalable, authentic localization – Coursera now delivers translated video content across multiple formats and languages without costly reshoots.
Mustafa encourages other teams to start small. ‘We began with one video. That’s all it took to see the impact,’ he said. ‘Once you experience it, it’s obvious what this can unlock.’
He also emphasised how fast AI technology, and HeyGen in particular, is evolving. 'This space moves quickly. If you’re not experimenting now, you’ll be behind later. HeyGen have become a critical part of our infrastructure.'
