Runway just shipped Aleph 2.0, an upgrade to its in-context video editing model, alongside a dedicated product surface called Edit Studio. The pitch is narrow but compelling: change exactly the pixels you want changed, and leave everything else identical to the source. No reshoots, no shot-by-shot babysitting, and no model hallucinating new cuts where none existed.

One frame in, a whole video out

The core workflow has shifted from prompt-only to image-first. A user can edit a frame to show the desired look, then have Aleph 2.0 carry that edit through the video. That reduces one of AI video's most expensive failure modes: waiting for a full generation just to discover that the model interpreted the request differently than intended.

Concretely, you pick a keyframe from your clip, modify it however you want (paint in a different shirt color, swap a product, drop in an object), and Aleph 2.0 treats that still as a visual spec. Instead of trying to describe complex visual modifications purely with text, you can upload just a single frame from your video that you have previously modified in any image editor. Aleph 2.0 will analyze this single keyframe and automatically project all visual modifications across the rest of the video sequence, handling motion tracking and perspective shifts on its own.