Overview
- Video Arena’s blind human-preference tests currently rank Gen 4.5 No. 1 among text-to-video systems.
- Runway claims gains in physics, human motion, camera movement and fluid dynamics to deliver more cinematic, photorealistic results.
- The company acknowledges lingering flaws including object permanence lapses and causal errors where effects can precede causes.
- Rollout is gradual, with access through Runway’s platform, API and select partners and full customer availability slated this week.
- The launch raises competitive pressure on Google and OpenAI, with Google’s Veo 3 at No. 2 and OpenAI’s Sora 2 Pro at No. 7, as Runway courts media and creative users.