Insider

Accelerating the Language Dubbing Process With NVIDIA-Powered AI Solutions

Objective

Insider uses Papercup's AI solution, which runs on NVIDIA A100 GPUs, to localize news and existing videos in a quick and cost-effective way, so it can reach and further engage with global audiences.

Customer

Insider

Partner

Papercup

Use Case

Media and Entertainment

Technology

NVIDIA A100 GPUs

Insider uses Papercup's AI solution, which runs on NVIDIA A100 GPUs, to localize news and existing videos in a quick and cost-effective way, so it can reach and further engage with global audiences.

Insider is an online global news publication that's part of Insider Inc., an American online media company. Since its acquisition by Axel Springer in 2015, Insider has steadily built a strong presence in English-speaking markets around the world. Next, the company wanted to extend that reach to international audiences -- specifically targeting non-English speaking markets where Insider video content was already showing traction.

Language dubbing replaces the dialogue in video content with translated dialogue tracks. This process is most commonly used for translating content into other languages as an alternative to subtitles and closed captions. However, traditional dubbing workflows are often expensive, and typically require large turnaround times, making it previously inaccessible to many companies.

Papercup is an AI startup on a mission to make the world’s videos watchable in any language. The company does this by automating large parts of dubbing, and using AI voices that sound human—all trained through their own machine learning algorithms. Papercup's AI solution, which runs on NVIDIA A100 GPUs, enables media companies, large corporations, and e-learning platforms to reach a global audience with their existing content.

With the help of Papercup, a member of the global NVIDIA Inception program for startups, Insider has been able to accelerate its growth in targeted markets by localizing its back catalog of informational video content.

A video by Insider, dubbed using Paperup's AI technology

Insider wanted to expand its reach to international audiences. Traditional language dubbing replaces dialogue in video content with translated tracks, but these workflows are often expensive and require large turnaround times. So Insider turned to Papercup’s AI solution, which runs on NVIDIA A100 GPUs:

Papercup's AI solution helps make videos watchable in any language, using AI voices trained through their own machine learning algorithms.
Papercup’s machine learning training workloads run on NVIDIA A100 GPUs, both on-premises and in the cloud.

With the help of Papercup, Insider has been able to accelerate its growth in targeted markets by localizing their back catalog of informational video content.
In the first 12 months of launching new channels for the LATAM and European countries on YouTube and Facebook, Insider was able to reach hundreds of millions of additional viewers.

The Need for Time-Saving Solutions to Localize Content

There were two important reasons Insider wanted to reach global audiences:

First, digital media is dependent on audience growth -- as measured primarily by reach and engagement -- to monetize its offering.
And second, offering audiences content that is in their native languages was a crucial step to bolstering Insider’s existing global brand equity.

But localizing news and factual content relies on faster turnaround times that align with the topicality of the content. This meant the team couldn't use traditional studio dubbing because it's often a time-consuming and expensive process.

Alternatively, Insider was using subtitles as the most cost-effective and speedy localization solution, but changing consumer habits means that engagement with dubbed content is higher. The partnership with Papercup has allowed Insider to dub its video content in a much more cost-effective way, with turnaround times that ensure the content remains relevant.

“... for underlying GPU compute, we can’t imagine running our training and inference workloads anywhere else. We see 5-10x training workload speedups out-of-the-box whenever we move to a newer NVIDIA architecture family."

James Leoni
Head of Machine Learning Papercup

Unlocking New Possibilities With Papercup, NVIDIA, and AI

Papercup’s AI-generated voices have a level of expressivity that outweighs the engagement generated by subtitling. The quality of the AI-generated voices is verified by human translators, which means that Insider can maintain its recognizable brand quality. And because AI does the heavy lifting, the localization process integrates with Insider’s existing content creation process, allowing the teams to localize with minimal effort, with the potential for high returns.

The deep learning training and inference workloads are accelerated by NVIDIA A100 GPUs. And voice synthesis in production takes place on NVIDIA GPUs in the cloud, managed by the Triton Inference Server. Moving their models in production to the Triton Inference Server is a key part of the strategy towards improved performance, flexibility, and utilization of GPUs during voice synthesis.

Snapshot of Papercup's AI dubbing platform during the QC process

Papercup's AI dubbing studio, automatically translating, segmenting and creating an AI voiceover

"Triton gives us the edge to scale our synthesis throughput and greatly eases optimization of our end-to-end latencies by supporting inference frameworks such as ONNX and TensorRT," said James Leoni, Head of Machine Learning at Papercup. "And for underlying GPU compute, we can’t imagine running our training and inference workloads anywhere else. We see 5-10X training workload speedups out-of-the-box whenever we move to a newer NVIDIA architecture family." The team at Papercup expects to see a large increase in GPU utilization in their inference clusters for lower cost, and reduced time to deploy a new model from hours into minutes.

Reaching New Audiences and Maximizing Engagement with Dubbed Content

Insider's global audiences are no longer hindered by language barriers— with the help of Papercup and NVIDIA-powered AI technologies, viewers can consume content in their native language. In the first 12 months of launching new channels for the LATAM and European countries on YouTube and Facebook, Insider was able to reach hundreds of millions of additional viewers—some of the localized videos even outperformed the English originals. In a matter of weeks, Insider gained 100 million views on its Spanish YouTube channel. "By localizing our existing content with Papercup, we’ve been able to unlock a whole new, Spanish-speaking audience for our existing content," said Tony Manfred, Head of Video at Insider. "Not only has this hugely increased the value of our content, it means we know more about how people consume it which is invaluable to future strategy. The future of media is dependent on its ability to reach and engage audiences and the Papercup solution directly improves the feasibility of hitting these key metrics. By improving the accessibility of content, not only does it increase the profitability of the content and longevity of the business, but it also gives audiences access to a greater pool of reliable news content.”