Pioneering New Interfaces in the Age of Generative Media
September 10, 2024
by Runway Team
Pioneering New Interfaces in the Age of Generative Media

We think AI is a new medium of creation, and like any medium, it demands fresh interfaces and novel grammars for interaction. Today, we want to share three prototype interfaces for generative media that can help shine a light on the possibility space as well as the challenges of this new medium. We believe in a future where most of the entertainment and media you enjoy and engage with will be generated in some form. The interfaces presented here are explorations designed for a generative media world.

While text prompts have dominated the landscape of media generation thus far, this is more a testament to their low barrier to entry than their expressive potential. The ubiquity of keyboards and the relative simplicity of text-based models have made language the de facto control mechanism for AI. But as we peer into the future, it's evident that text alone won't suffice as the primary conduit for our creative dialogue with AI.

The interfaces of tomorrow will be as diverse and dynamic as the content they help create. There are a few design principles that inform our interface explorations, and four fundamental design principles we tend to follow: wonder, discovery, control, and feedback.

Wonder and Discovery

The most powerful creative tools don't just execute commands - they inspire. Our interfaces should be playgrounds for the imagination, spaces where serendipity and exploration are not just allowed but encouraged. We're designing systems that facilitate "generative daydreaming," where the journey of creation is as vital as the destination. Often, the most profound creative breakthroughs occur when we're not sure where we're going, but the tools are opening the space to wonder and experiment.

Concept Cluster: Suggested groups of creations or assets based on a simple conceptual prompt presents a holistic vision of new possibilities.
Control

While we celebrate the unexpected in the creative process, we also recognize the need for precision. Future interfaces will offer granular control, allowing creators to fine-tune their visions at the pixel level. But this control goes beyond mere adjustment - it's about providing clear mechanisms for interacting with variations, for diving into the latent space of possibilities and steering the AI's output in meaningful ways. Imagine being able to reach into a generated image and manipulate its underlying structure, or to guide the evolution of a story by tugging on narrative threads.

Dynamic Control: Dynamically generated controls based on a prompt can expand creativity, and increase control
Feedback

Creation is a dialogue, not a monologue. Interfaces should provide robust, real-time feedback mechanisms that turn the act of creation into a true conversation. You do something, you see something. This immediate responsiveness changes the direction and interestingness of the creative space, allowing for rapid iteration and exploration. A virtuous cycle of inspiration and refinement, where each action opens up new avenues of possibility.



Mashup concept: Combine images to get outputs that you either expect or didn’t expect.
Snap Edit: Intuitive controls to edit your generations.

It's uncharted territory and we're acutely aware of the challenges that lie ahead. How do we balance intuitiveness with power? How do we create interfaces that are accessible to novices yet deep enough for experts? How do we ensure that these new tools enhance human creativity rather than supplant it? These are the questions that drive our exploration. Exploring new ways of thinking about and interacting with machine intelligence.

Title
Pioneering New Interfaces in the Age of Generative Media
Authors
Edwin Toh, Steph Jeong, Alejandro Matamala-Ortiz
Year
2024
Runway 2024 | All rights reserved.