Synthesia Collaborates With Shutterstock, Trains EXPRESS-2 AI Model To Create Realistic Avatars

JAKARTA - Synthesia, a technology company focused on developing artificial intelligence-based videos (AI), announced its partnership with Shutterstock, a digital content provider.

Through this partnership, Synthesia will use the extensive Shutterstock library to pre-train their newest AI model, EXPRESS-2, which will support a new generation of AI avatars later this year.

As part of the agreement, the Synthesia R&D team will have access to Shutterstock's content library through a research license, with the option to expand it for commercial use.

Meanwhile, company AI researchers will review the high-quality HD and 4K video data provided by Shutterstock and select relevant content to perfect their existing datasets.

They admit that building a large AI model will involve many experiments focused on research before it can be produced. This is done to prove that their AI model becomes more sophisticated.

"Since last year, we have tested various approaches and architectures for pre-exPRESS-2 training, our second attempt at building a large video and audio model for our platform," wrote Synthesia's Head of Corporate Affairs and Policy, Alexandru Voica in its official blog.

This is expected to be able to create an AI model that is able to reproduce its natural and realistic behavior, movement, and expression, such as delivering scripts with the right tone and with facial expressions as well as correct body language movements.

Synthesia has even started this research project in 2023, where they paid thousands of actors and studied their performances in studios around the world.

"Thanks to this partnership with Shutterstock, we hope to try a new approach that will improve EXPRESS-2 performance, and improve the realism and expression of the avatar we generate with AI, so that it is closer to human-like performances," he concluded.