Synthesia unveils HumanRF: a leap in ultra-realistic AI-generated human movement

humanrf renders

AI talking head video creation company Synthesia has released a new high-performance system for creating ultra-realistic human movement playback scenes.

Hailed as a “huge announcement” in the AI world by Rowan Cheung, who is behind influential AI newsletter The Rundown, the system is called HumanRF and it’s claimed that it enables playback from “unseen, novel viewpoints”, meaning it can create video from viewpoints it didn’t directly film.

Synthesia said that the aim of the development was to “close the gap to production-level quality” of digital human representation for uses such as in movies, video games and videoconferencing. They didn’t mention sextech purposes, but anyone who’s been to a ‘virtual stripclub’ that looks like a mildly erotic version of The Sims will see the potential.

humanrf full body multi view vid

HumanRF records human actors with a set-up including 160 cameras, filming at 12MP resolution, far higher than most render recordings in the field. It uses processes such as volumetric rendering to create 3D renderings out of the areas of the human only captured in 2D, offering a ‘complete’ digital rendering that can be viewed from any angle.

The result is basically an extremely impressive digital human rendering that remains realistic even when the human render is performing fast actions, such as dancing. To minimize blurring when fast actions such as these were filmed, the researchers set their cameras to have shutter speeds of 650us.

Digital 3D human renders have been around for years in film, video games and elsewhere, but the use of AI to create these ‘complete’ angle renders, plus the leap forward in resolution and movement quality, feels like a significant step forward. Makers of erotic augmented reality (AR) services will also be watching carefully.

Synthesia said that by launching HumanRF, you can now create “temporally coherent reconstructions of human actors for long sequences, while representing high-resolution details even in the context of challenging motion.”

Affiliate Disclosure
Some articles contain affiliate links that allow us to earn money if you decide to purchase any of these products or services. This does not cost you any extra money, and it allows us to continue to run this website. Affiliate links have no relation to review ratings or other editorial coverage. You can read the full policy here.

Jamie F

Jamie F

Jamie is a freelance writer, contributing to outlets such as The Guardian, The Times, The Telegraph, CNN and Vice.

Be the first to leave a comment

Leave a reply