Insider needed a faster and more cost-effective way to localize video content for non-English-speaking audiences because traditional dubbing was too slow and expensive and subtitles delivered lower engagement.
Reported impact
Insider reached hundreds of millions of additional viewers, gained 100 million views on its Spanish YouTube channel within weeks, and expanded international audience engagement with dubbed content that sometimes outperformed the English originals.
Insider localized its video catalog by plugging Papercup's AI dubbing platform into existing publishing workflows instead of relying on slow studio dubbing or subtitles alone. Papercup trains and serves its speech models on NVIDIA A100 GPUs, using GPU acceleration for both model training and production voice synthesis. As clips move through the workflow, Papercup automatically translates dialogue, segments audio, generates human-sounding AI voice tracks, and routes the output through quality-control steps with human translators before release. NVIDIA Triton Inference Server is being adopted to raise synthesis throughput, simplify support for inference frameworks such as ONNX and TensorRT, and improve end-to-end latency and GPU utilization. This architecture let Insider dub back-catalog news videos quickly enough to keep content relevant while expanding distribution into Spanish, Latin American, and European channels.
Existing Insider videos enter Papercup's dubbing workflow, dialogue is translated and segmented, AI voices are generated on NVIDIA GPU infrastructure, human translators verify quality, and localized videos are published to regional channels.
In a matter of weeks, Insider gained 100 million views on its Spanish YouTube channel.
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.
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.
Papercup's AI solution helps make videos watchable in any language, using AI voices trained through their own machine learning algorithms.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Sep 14, 2026 Pinterest Pinterest needed to serve vision-language AI for shopping and discovery at internet scale, but multimodal requests involving thousands of images created prefill-heavy inference, memory pressure, and routing challenges… 14,000 NVIDIA GPUs AI Configurator · Dynamo Planner · NVIDIA Blackwell B200 GPUs NVIDIA Sample Aug 25, 2026 Neuron Systems Neuron Systems needed an always-on AI sports newsroom that could analyze every play, sustain multi-agent reasoning during long live broadcasts, and generate multilingual commentary without the latency, rate limits, and… 1 million audited AI traces NVIDIA CUDA · NVIDIA H100 GPUs NVIDIA Sample Aug 25, 2026 PYLER PYLER needed to verify brand safety across millions of videos per day, but keyword and rule-based moderation could not provide the semantic precision, throughput, or operating cost profile required for real-time… 10X reduction in operational costs NVIDIA CUDA · NVIDIA DGX Platform · NVIDIA Mission Control 