.footer { } Logo Logo
deutsch
/// News
Resolve 18.5: New features explained with Screencaptures: Relight, Text-Based Editing ...

Runway Gen2: New text-to-video AI from the Stable Diffusion creators

[19:58 Mon,20.March 2023   by Thomas Richter]    

Runway Research, the company behind the image AI Stable Diffusion, has just announced the new video AI Runway Gen2, whose capabilities significantly exceed those of the first version (Runway Gen1). While this was still limited to changes to an existing video (video-to-video) via prompt or pre-image, the new version now finally offers the ability to generate arbitrary videos via text prompt only (text-to-video) - quite analogous to the generation of images using the text-to-image AIs such as Stable Diffusion or Midjourney. Alternatively, individual images can be brought to life in the new variation mode (image-to-video).



Runway-Gen2-examples
Runway Gen 2 sample clips



Prompt: "Aerial drone footage of a mountain range"


Runway Gen 2 builds on the features of Gen 1, so in addition to generating video via text prompt, it also offers its features, such as the ability to manipulate the style, composition, or even content of a video based on a text prompt or sample image (similar to the recently shown but less sophisticated Tune-a-Video).



In addition, a special stylization mode can be used to change the style or look of a clip via text description, while the storyboard mode can be used to convert a simple moving sketch into a finished animation. A sixth mode provides masking functionality - objects can be selected by text, for example, and modified selectively. In the seventh mode, rough 3D models can be supplemented with suitable text textures to create rendered objects. And finally, in the customization mode, it is possible to adapt Gen-2 specifically to one&s own needs.

Prompt: "The late afternoon sun peeking through the window of a New York City loft."



The Runway Gen 2 video also demonstrates how steadily Gen 1 image quality has improved since launch. And indeed, the videos generated via Gen 2 are of very high quality: the otherwise notorious inconsistency of subjects from frame to frame has been improved, as has the quality of composition and general rendering, which now has very few image errors.

Runway-Gen2-evolution
Improvements in Runway Gen 1 quality over the course of a few weeks.



Compared to VideoFusion, the just released first video AI whose source code including models is published and which also runs on home PCs, the quality (and resolution) of Gen 2 is already several generations ahead. But even the comparison to the video AIs of the big players like Google (Imagen and Phenaki ) and Meta ( Make-a-Video) need not shy away from Runway Gen 2. Runway Gen 1 is currently only available in closed beta form, Gen 2 is scheduled to launch soon.

Runway-Gen2-quality


Link more infos at bei research.runwayml.com

deutsche Version dieser Seite: Runway Gen2: Stable Diffusion Schöpfer stellen neue Text-to-Video-KI vor

  



[nach oben]












Archiv Newsmeldungen

2024

July - June - May - April - March - February - January

2023
December - November - October - September - August - July - June - May - April - March - February - January

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000






































deutsche Version dieser Seite: Runway Gen2: Stable Diffusion Schöpfer stellen neue Text-to-Video-KI vor



last update : 26.Juli 2024 - 18:02 - slashCAM is a project by channelunit GmbH- mail : slashcam@--antispam:7465--slashcam.de - deutsche Version