Runway's Latest AI Video Generator Brings Giant Cotton Candy Monsters to Life

Enlarge / Screen capture of A Track Generation 3 Alpha video generated with THE fast "A giant humanoid, do of fluffy blue cotton candy, trample on THE ground, And roar has THE sky, clear blue sky behind them." Track

On Sunday, Track announcement A new AI video synthesis model called Generation 3 Alpha It is always below development, but he appears has create video of similar quality has OpenAI Sorah, which made his debut earlier This year (And has Also not Again has been released). He can generate novel, HD video Since text instructions that range Since realistic humans has surrealist monsters trample THE campaign.

Contrary to Track previous best model Since June 2023, which could only create duration of two seconds clips, Generation 3 Alpha can would have create 10 seconds video segments of people, places, And things that to have A consistency And consistency that easily exceeds Generation 2. If ten seconds sounds short compared with has Sora's complete minute of video, consider that THE business East functioning with A limited means budget of calculate compared with has more sumptuously finance OpenAI—and In fact has A history of shipping video generation aptitude has commercial users.

Generation 3 Alpha do not generate audio has to accompany THE video clips, And It is very likely that temporally consistent generations (those that keep A character consistent on time) are addicted on similar high quality training material. But Track improvement In visual loyalty on THE pass year East difficult has ignore.

AI video heat up

It is has been A busy couple of weeks For AI video synthesis In THE AI research community, including THE launch of THE Chinese model Kling, created by Based in Beijing Kuaishou Technology (Sometimes called "Kwai"). Kling can generate two minutes of 1080p HD video has 30 executives by second with A level of detail And consistency that would have matches Sora.

Generation 3 Alpha fast: "Subtle reflections of A women on THE window of A form moving has hyper-speed In A Japanese city."

Not long After Kling made his debut, people on social media began create surrealist AI videos using Lumia AI Lumia Dream Machine. These videos were novel And weird but in general missing consistency; We tested out Dream Machine And were not impressed by Nothing We saw.

In the meantime, A of THE original text to video pioneers, New York City based Track—founded In 2018—recently find himself THE end of memes that watch It is Generation 2 technology fall out of service compared with has more recent video synthesis models. That can to have stimulated THE announcement of Generation 3 Alpha.

Generation 3 Alpha fast: "A astronaut running through A alley In Rio of Janeiro."

Generator realistic humans has always has been difficult For video synthesis models, SO Track specifically watch disabled Generation 3 Alpha ability has create What It is developers call "expressive" human characters ...