Runway unveils "Gen-4" — video generation that keeps characters and locations consistent
Runway announced its latest video generation model, “Gen-4.” Its standout feature is reference conditioning using reference images, which lets characters, objects, and environments stay visually consistent across multiple video generations without any additional fine-tuning.
Details
- Achieving consistency: Realizes so-called “world consistency,” letting the same characters, locations, and objects appear across separate generations while keeping their appearance intact
- Output specs: Generates 5- or 10-second clips at up to 720p, with 4K upscaling available on paid plans. Frame rate is 24fps
- Follow-up model: Just days after the announcement, Runway also released a faster version, “Gen-4 Turbo,” which boosts generation speed roughly fivefold
- Aim: Runway positioned this as a step toward a “world model” that simulates the physical laws of the real world, and said motion dynamics had also improved compared to Gen-3 Alpha
What happened next
Gen-4’s character consistency gave a boost to AI adoption in live-action film and animation production. Just three days after the announcement, on April 3, Runway announced a $308 million Series D funding round, which went on to fuel the expansion of its film and animation production division, Runway Studios.