Aleph 2.0 is an in-context AI video editing model developed by Runway, released in May 2026. Unlike traditional text-to-video models that generate footage from scratch, Aleph 2.0 is designed to transform and modify existing video clips while preserving the original motion, timing, and continuity. It functions as the core engine for Runway's Edit Studio, allowing users to perform complex post-production tasks such as object swapping, relighting, and scene restyling through natural language prompts and visual guidance.
The model introduces a workflow centered on single-frame editing, where users modify a specific frame in a video to serve as a visual target. Aleph 2.0 then propagates those changes across the entire clip, maintaining consistency in lighting, shadows, and reflections. It supports high-definition output at 1080p for clips up to 30 seconds in length, representing a significant extension over the original Aleph model's 5-second capacity.
Technically, Aleph 2.0 is built to understand the spatial and temporal context of a scene. This enables it to handle multi-shot sequences and maintain visual fidelity across cuts and camera angle changes. The model can isolate specific subjects for localized edits, such as changing a character's wardrobe or updating product branding, without affecting the surrounding environment or the actor's natural movements.
For optimal results, users are encouraged to use frame anchors—pinned reference images that guide the model at specific timestamps. When prompting, it is recommended to describe only the elements that need to change, as the model's primary objective is to keep unmentioned parts of the source video stable. The system supports up to five anchored reference images to ensure precision in long-form edits.