It feels like every few weeks we see a new image generation model roll-out, but it's not often one of AI's major video creation players drops one — and that's exactly what has happened with Runway AI.
The company has introduced 'Frames', its newest base model for image generation. Runway says the new image model marks "a big step forward in stylistic control and visual fidelity".
There's a focus on integrating stylistic choices and more reliably making it consistent throughout the model, something many models have struggled with. This in turn allows for more consistency across video generations using the same 'world'.
Runway's new image generation model is rolling out
Introducing Frames: An image generation model offering unprecedented stylistic control.Frames is our newest foundation model for image generation, marking a big step forward in stylistic control and visual fidelity. With Frames, you can begin to architect worlds that represent… pic.twitter.com/kg8Fz0LEgUNovember 25, 2024
In a new blog post, the team at Runway says it's rolling out access to Frames inside Gen-3 Alpha and the Runway API 'gradually'. We spoke to the company and it said it would happen over the coming weeks.
You can see some examples of impressive usage of Frames in the X thread above, from 1980s SFX-inspired makeup that wouldn't have looked out of place in Labyrinth, to 1970s album art and more. There was even one 'world' shared by Runway's Head of Creative Jamie Umpherson on X where everyone has a duck head.
In fact, I'm hard-pressed to point out any obvious tells that the models generated aren't real humans — and that's not something I can say of every model.
I'm particularly impressed with the Japanese Zine look which adds the kind of style over accuracy that can often flummox other models — it's almost too stylized for other models, but Runway's Frames looks great.
