Vyond Reduces Generative AI Video Costs by 20% Using AWS Inferentia

In 2021, Vyond—a global video creation platform based in the US with offices in Hong Kong and Taiwan—secured $50 million in growth capital to enhance its position in the enterprise video creation market. The company helps businesses craft high-quality animated videos, fostering engagement and education among employees and customers.

Vyond’s flagship offering is Vyond Studio, a drag-and-drop application that utilizes artificial intelligence (AI) for efficient video creation. In early 2023, the company sought to extend its product portfolio with a generative AI–based application that could create high-quality videos with a few text prompts. “We saw how generative AI models could produce unique stylistic videos with just a few words,” explains Matt Harney, chief technology officer at Vyond.

Vyond required infrastructure with the computing power to host deep learning, text-to-image models, and the flexibility to scale quickly as demand for the new service grew. The company turned to Amazon Web Services (AWS), which was already providing the infrastructure for Vyond Studio. “We knew AWS could help us innovate quickly and keep pace with advancements in generative AI technology,” says Harney. “AWS offered the elasticity we needed, plus access to the latest-generation GPUs driving today’s AI developments.”

Vyond replicated the AWS architecture used in Vyond Studio when implementing Vyond Go. To enhance scalability and cost efficiency, the company utilized Amazon Elastic Compute Cloud (Amazon EC2) Inf2 instances to host several machine learning (ML) models, including Stable Diffusion, a generative AI model that produces unique stylistic images from text and image prompts. In addition, Vyond deployed Amazon EC2 Inf1 instances for other AI-powered features, such as visual content search, video-to-action, and background removal. Says Harney, “By adopting Amazon EC2 Inf2 instances for our generative AI workloads, we saw a saving of 20 percent.”

Vyond also implemented Amazon Elastic Kubernetes Service (Amazon EKS), a managed Kubernetes service, for automated application deployment, scaling, and administration, and for heavy server workloads to reduce the overall compute cost.

For ongoing optimization, the company works with AWS Enterprise Support, including a technical account manager, who provides operational assistance and development guidance.  “That’s the advantage of working with AWS,” says Harney. “We get great support on the best way to run our different AI workloads in the cloud, so we’re constantly improving our infrastructure.”

To maximize operational performance, Vyond is also using Amazon OpenSearch Service for logging and utilizing Amazon Managed Service for Prometheus, as well as Amazon Managed Grafana for monitoring and maintaining consistent response times for its large container deployment.

By running on AWS, Vyond can ensure that Vyond Go meets stringent enterprise-level security standards. This involves compliance with ISO 27001, GDPR (General Data Protection Regulation), CCPA (California Consumer Privacy Act), and FedRAMP (Federal Risk and Authorization Management Program) certification for US government agency business. Fei Yu, senior research and development manager at Vyond, adds, “Using AWS, we can support all the latest advancements in AI while ensuring optimal security, so business customers get the compliance protection they need.”