Runway previews Continuum video generation from screen selections

• Continuum turns screen selections into generated video without typed prompts, according to Runway's preview.
• Window resizing can change scene content: Runway's experiments rearrange boxes and turn a ski map into a slope-level view.
• A drawing prototype animates sketches while retaining their visual style, as Runway demonstrates.

The full episode includes a ski map that shrinks into one slope and hand-drawn eyes that start looking around. Runway Labs designer and prototyper Ryan Stephen presents the experiments in Runway's latest inside view, which the company frames around learning, creation and collaboration.

Continuum explores real-time video as part of an operating system. Stephen describes the interaction in his walkthrough:

• Starting point: a webpage, a video or other content already in view.
• Gesture: select an area of the on-screen canvas.
• Output: generate video using the existing context, without writing a text prompt.

The screen-selection gesture is the standout idea here. Stephen describes flexible learning experiences as one possible use, with the generated video helping someone explore what they are already viewing.

The preview gives no public access link or release date for Continuum.

Stephen says the model can reinterpret a scene for a smaller frame. The experiments give window dimensions a role in determining what the video contains:

• Moving boxes: the boxes reshuffle dynamically as the scene is resized.
• Ski map: shrinking the video turns a broad map into a richer view of a single slope.

The tablet drawing prototype has a specific update rhythm: a real-time video model interprets the drawing every few seconds, Stephen explains in the demonstration. It adds motion while keeping the artist's style.

• Cartoon eyes: simple circles become eyes that blink and look around.
• Office experiments: Stephen says colleagues also drew butterflies, complex scenes and memes.