Hot on the heels of a new version Midjourney and the launch of GPT-4, Runway, whose web-based video editor specialises in background removal and pose detection and who helped to develop open-source text-to-image model Stable Diffusion, has trailed the imminent arrival of Gen-2, a new version of its text-to-video tool.
Gen-1 allows users to amend existing video footage by overlaying rough AI-generated 3D animations or shaky smartphone clips as overlays. Get-2 is going to be more about generating new videos from scratch. Early results are short, unstable and in no way photorealistic.
Text-to-video research from Meta and Google is more advanced that Runway’s, with longer, more cohesive clips, but then again they have considerably larger resources at their disposal – Runway’s team numbers 45.
If you are interested in using generative AI and other forms of machine learning in your creative processes, Mondatum provides a range of services around the development of artificial neural networks (ANNs), machine learning model training and the use of AI to solve problems and streamline processes and workflows. For more information, contact Colin Birch – colin@mondatum.com.
Source: The Verge