Runway unveils "Gen-4.5" β takes the top spot on text-to-video benchmarks worldwide
Runway announced its latest video generation model, βGen-4.5.β It says the model scored an Elo of 1,247 on Artificial Analysisβs published text-to-video benchmark, taking first place ahead of models from Google and OpenAI.
Details
- Benchmark results: Ranked first in the world on the Artificial Analysis Text to Video benchmark, with an Elo score of 1,247
- Physical representation: Achieves more realistic physics simulation, including object momentum and liquid behavior
- Detail consistency: Fine details such as individual strands of hair and material textures now hold together better through motion
- Range of styles: Supports a wide range of styles, from photorealism to stylized animation
- Character expression: Improved emotional expression and facial nuance enable more natural character performances
How to try it
- Gen-4.5 is available at no extra cost on all existing Runway paid subscription plans
- Existing control modes such as Image to Video, Keyframes, and Video to Video are being rolled out to support Gen-4.5 as well
- Enterprises can also inquire about custom solutions