The Starry Night at University Library

A message from the Dean

Happy Holidays!
Thank you for visiting this website and being interested in the application of AI to create the SJSU Library's 2023 holiday card. This card is different from past holiday cards. Typically, libraries (including ours) highlight items from our collections (a wonderful and important tradition). However, this year we decided to go in a different direction with this card. Libraries are at the forefront of technical advances and discovery. Since AI has really jumped into the forefront of so many conversations, this holiday card highlights the library's creativity and innovative mindset. I'm very grateful to the library team, in particular Lisa, Jon, and Ray, who all turned a spark of an idea into this wonderful design. Along the way our library marketing team and their outstanding student team added another touch and extra pizazz to the card. I hope you enjoy the card as much as we did creating it.
If you have any comments or questions, we'd love to hear from you.
I wish you and yours a joyous and rejuvenating holiday period!
Here's to a great 2024.

- Michael Meth - Dean, University Library

A photograph of Dr. Martin Luther King, Jr. Library taken by SJSU King Library Student Assistant Maleek Diaz.

Michael Meth
Dean, University Library

The story behind the Artificial Intelligence assisted design of SJSU King Library’s 2023 Holiday card

The team, led by Dean Dr. Martin Luther King Jr. Mike Meth, received a directive to create artwork that symbolizes the forward-looking nature of SJSU King Library. During a collaborative brainstorming session, the decision emerged to explore the innovative concept of generating artwork through Artificial Intelligence (AI).

To guide this endeavor, Jon Oakes, the Technology Labs Coordinator at SJSU King Library, was enlisted to provide expertise on AI image generation models. His role involved advising on the functionality of AI models and making informed choices aligned with ethical considerations. The selection of Style GAN was primarily influenced by the fact that the generated results would adhere to Van Gogh’s artistic style, and, crucially, Van Gogh's works were part of the public domain.

A photograph of Dr. Martin Luther King, Jr. Library taken by SJSU King Library Student Assistant Maleek Diaz.

Original photograph of Dr. Martin Luther King, Jr. Library taken by MLK Student Assistant Maleek Diaz, Political Science 2025

What is a Style Generative Adversarial Network?

Created by researchers at Nvidia, Style Generative Adversarial Network is a machine learning framework. The AI Discriminator refers to Van Gogh’s library of paintings as a benchmark to measure the results produced by the AI image generator. Under the guidance of the AI Discriminator, the AI image generator iterates millions of results until a result satisifies the benchmark set by the AI Discriminator.

Infographic describing how the Artificial Intelligence Discriminator refers to Van Gogh’s library of paintings as a benchmark to measure the results produced by the AI image generator.

The Artificial Intelligence Discriminator refers to Van Gogh’s library of paintings as a benchmark to measure the results produced by the AI image generator.

Style GAN in action

Infographic describing how the Artificial Intelligence Discriminator refers to Van Gogh’s library of paintings as a benchmark to measure the results produced by the AI image generator.
  1. The next phase in the project was uploading the photograph of Dr. Martin Luther King, Jr. Library taken by Maleek Diaz to the AI Image Generator.

  2. Given the prompt “library in the art style of Van Gogh’s The Starry Night”, and under the guidance of the AI Discriminator, the AI Image Generator begins producing millions of iterations of the photograph provided.

  3. A result is produced by the AI Image Generator and is rejected by the AI Discriminator.

  4. Another result is produced by the AI Image Generator and is rejected by the AI Discriminator.

  5. Learning from the millions of results and rejections, the AI Image Generator produces an iteration that satisfies the benchmark set by the AI Discriminator.

Style GAN results

The four results below were chosen as candidates for a staff wide voting process. Candidates 1 and 3 received the most votes.

Adobe Photoshop and the final touches

Staff members requested that the sky from Candidate 1 be combined with the building from Candidate 3. This composite was created in Photoshop by SJSU King Library Adobe Trainer & Instructional Designer Raymond Lam, who used techniques like masking, color correcting, as well as blending in paint strokes to give the image a more painterly texture and feel.

Timelapse video of the Adobe Photoshop compositing process sped up by 400%.

It’s Time to Design!

Pulling inspiration from the AI generated image, the Library Marketing Team (LMT) came together to design a holiday card and a digital copy with motion and animations. Using Adobe Illustrator, LMT decided to extend the image to the back cover with the continuous swirl, giving it the illusion of motion while also staying true to the original image.

From there the team extracted the blues from the skyline for the background and text to give the overall design a cohesive look and feel. With a touch of ‘Spartan Pride’ (font), the card was officially completed and arrived straight to your mailbox! 💌

Bringing It to Life with After Effects

Staying true to the original design and the natural movements of the skyline, an animated version of the card was created using Adobe After Effects. Moving from left to right and orbiting around the moon gives the illusion of a gusty, winter wind. 🌬️

Final thoughts

“In the past year, AI has advanced rapidly, prompting us to consider ethical and responsible ways to harness its power. Adopting new technology without careful consideration poses risks, so we are actively exploring and experimenting, encouraging mutual challenges within our team. Despite initial reservations, the capabilities of AI surpassed our manual efforts, resulting in a surprisingly successful outcome. We are confident that our use of AI technology in this instance was not only efficient but also respectful of the source material and the contributions of the artists involved, ultimately producing a high-quality product.”

- Jon Oakes, Dr. Martin Luther King, Jr. Library Technology Labs Coordinator

Contributors

Michael Meth

University Library Dean

Lisa Josefik

Executive Assistant to the Dean

Lesley Seacrist

Project & Communications Manager

Mariah Ramsour

Communications & Events Coordinator

Jon Oakes

Technology Labs Coordinator

Raymond Lam

Adobe Trainer and Instructional Designer

Janhvi Kriplani

'24 MS Human-Computer Interaction

Joseph Shin

'24 BA Graphic Design

Maleek Diaz

'25 Political Science

Van Nguyen

'23 BFA Digital Media Art