This is the category that makes Hollywood sweat. Rightfully, because video GenAI has the potential to extinct half of the film industry. If e.g. Runway or Sora AI allows creating any scenery (set) 100% according to the director's vision on the computer screen, they will prefer that over sending a costly film crew around the world.
We're not there yet, for now the AI video generators don't follow prompts well enough. Too many generations are needed to make something decent, but we're slowly getting there. For example this music video by Paul Trillo was made entirely with AI video.
It took them 8000 generations with SoraAI and according to Paul Trillo getting an evenly motion pace was the most challenging part.
To give you more of an overview of where we stand currently with video GenAI, aka AI video generators, let me introduce the main players.
Runway is the most famous tool and so far still the front runner. SoraAI was very much hyped when OpenAI first introduced the tool, but couldn't quite keep up with the expectations once it hit the market. Google's VEO 2 has surprised some in the video gen community. Pika labs was also one of the first video software to create a entirely new video sequences from a prompt, but seems to be doing best with animation style.
In the years coming, expect a fierce battle of the mentioned main players. Video generators will excel in making less mistakes, more relatable characters and near undistinguishable sets and sceneries from real ones.
Also, check out or other two video related categories AI video editors and AI Avatars.
