Don't miss leaders from OpenAI, Chevron, Nvidia, Kaiser Permanente, and Capital One, only at VentureBeat Transform 2024. Gain key insights about GenAI and grow your network at this exclusive three-day event. Learn more
RunwayML, one of the first startups to tackle the challenge of training AI for video generation, today announced the general availability of its latest Frontier model, Gen-3 Alpha.
First announced a few weeks ago, Gen-3 Alpha enables users to create hyper-realistic AI videos from text, image or video prompts. Broad access begins today, enabling anyone who subscribes to the RunwayML platform to use the model's high fidelity and controllable generation to enable a variety of creative use cases, including advertising, like the one OpenAI previewed with Sora.
But there's a catch: Gen-3 Alpha isn't free like the Gen-1 and Gen-2 models — users will need to upgrade to one of the company's paid plans, which start at $12 per editor per month, billed annually.
What can we expect from Gen-3 Alpha?
After releasing its Gen-1 and Gen-2 models a few months apart last year, RunwayML didn't announce any model releases and continued to push feature updates to its platform. During this period, several competitors showed off their wares, including Stability AI, OpenAI Pika, and most recently Luma Labs.
Countdown to VB Transform 2024
Join enterprise leaders at our flagship AI event in San Francisco July 9-11. Network with your peers, explore the opportunities and challenges of generative AI, and learn how to integrate AI applications in your industry. Register now
As the AI video wars gather pace, the startup decided to relaunch with Gen-3 Alpha last month. Trained on highly detailed captioned videos and images, the model enables users to produce ultra-realistic video clips featuring imaginative transitions, precise keyframing of elements, and expressive human characters exhibiting a wide range of actions, gestures, and emotions.
The samples shared by Runway show just how big of an upgrade Gen-3 Alpha is over Gen-1 and Gen-2 in terms of speed, fidelity, consistency, and movement. The company said it partnered with “an interdisciplinary team of research scientists, engineers, and artists” to develop these features, but didn't reveal much about the exact source of the training data.
Now that Gen-3 Alpha is broadly available, users can upgrade to Runway's paid plans and use it for video production for a variety of creative use cases.
Runway says that initially, the model will be used for text-to-video mode, allowing users to bring their ideas to life with natural language prompts. But in the coming days, it will be applied to other modes, including image-to-video and video-to-video. The model also works with the control features of the Runway platform, including motion brushes, advanced camera controls, and director mode.
Each video generated by the model is up to 10 seconds long, with the generation speed varying depending on the length of the video, which is better than most AI video models but falls short of the one-minute generation that OpenAI has promised with Sora (which has not yet been released).
It will be interesting to see how the creative community utilizes Gen-3 Alpha, and Emad Mostaque, former CEO of Stability AI, has already tried out Gen-3 Alpha to see how it compares to Sora's generation.
It's important to keep in mind that this is just the beginning, however: Gen-3 Alpha is expected to improve over the coming months, and a free version of the model will also be made available to users at some point.
Runway describes it as “the first in a series of upcoming models” trained on its new infrastructure built for large-scale multi-modal training, and as “a step towards building general-world models capable of representing and simulating a wide range of situations and interactions that we might encounter in the real world.”
Source link
