reimagineAI Playground

AI-Play of the month: GauGAN2

Creating artificial landscape photographs from scratch

For our first Play of the new year, we have chosen an updated demonstration of GauGAN: GauGAN2. If you are new to our monthly articles on AI in practice, make sure to check out the 25+ demos we featured last year.
As an image-generator, the web demonstration by US tech company Nvidia “is one of the first to combine multiple modalities — text, semantic segmentation, sketch and style — within a single GAN framework”. In GauGAN2 you have various options to create your own real-looking but artificially-build landscapes by using digital painting tools, natural language as well as your own images and a range of pre-defined filtering functions. See how it looks like in practice in the video below.


By loading the video, you agree to YouTube’s privacy policy.
Learn more

Load video

What’s GAN?

GAN is short for generative adversarial network, a popular class of machine learning. Those deep generative models such as GANs are characterized by their “ability to synthesize endless realistic, diverse, and novel content with minimal user effort. The potential utility of these models continues to grow thanks to the increased quality and resolution of large-scale generative models in recent years” as Wang et al. (2021) write in their paper . To complete the origin of the word-composition: The Gau stands, as it were, as a homage representative for Paul Gauguin, a French Post-Impressionist artist born in 1848 who became influential in the art world postmortem.

The image on the left was originally a quick drawing. After the image-generation it becomes possible to overlay the original one with texture mapping.

Lack of usability ≠ lack of power

Even though the website where the demo is programmed on looks like an old Paint interface from 2000, it becomes impressive when playing with the model and adjusting the output. It might take a second to understand what is actually happening but thereafter the fun begins when generating and transforming artificial landscapes based on your basic and trivial sketches. Since the model was trained on 10 million high-quality landscape photos you can be sure to create unique pictures.

Screenshot of a pre-given set of different styles.

In the future, generators such as GauGAN2 can be a great starting point for art generation, or for capturing moods and then visualizing them. They open-up chances to enhance and increase human creativity and productivity through the addition of AI. Further, the texture mapping developments and classifications will be quite helpful for visual work.

Every month we present two new applications for our GEDANKENFABRIK AI Playground. Read the intros, visit our Playground and try them out yourself!

Head over to our Playground!


AI-Plays of the month: GPT-3 & Codex
AI-Plays of the month: Ask Delphi & Deep Dream
AI-Plays of the month: Stealing Ur Feelings & air-drawing
AI-Plays of the month: AI21 Studio & Wordtune
AI-Plays of the month: Zendo & Scroobly
AI-Plays of the month: JFK Files & Gen Studio
AI-Plays of the month: The Moral Machine & Interactive TV
AI-Plays of the month: Sythesia & Giorgio Cam
AI-Plays of the month: Deep Nostalgia & AI Dungeon
How to play: First time experiences
Time to Play: The AI Playground