Introducing ImagePlot 2 — a new version of the ImagePlot tool we developed in the early 2010s.
- Input: a set of images and a metadata file (with a list of filenames and any other info you want to add, such as dates when artworks were created, artists' names, etc).
- The tool automatically extracts 48 visual features: brightness, saturation, hue, contrast, entropy, and more.
- Plot images as a scatter chart using computed features and the metadata,
- Compute PCA / t-SNE to visualize feature space structures.
- Quickly create multiple views using the same image set.
- Animate your visualization - for example, to see changes over time.
- Save any visualization as a .png.
The new ImagePlot 2 runs in a single HTML file — no installation, no server, no account.
Just open in a browser.
HTML file, user guide, and two sample image collections (van Gogh's 776 paintings, and 128 Mondrian's paintings made in 1905-1918) - link in a bio, or download it directly from here:
/5bdfp976
Our book 'Artificial Aesthetics: Generative AI, Art and Visual Media'—previously released chapter by chapter—is now available as a single free PDF file:
/index.php/projects/artificial-aesthetics
Translations into Spanish and Chinese are already in progress and expected later this year. If you work with a publisher who may be interested in translating the book into other languages, please contact me directly. We are also open to working with an English-language publisher, if they can move quickly.
***
Our book uses approaches from eight fields. Emanuele brings perspectives from aesthetics, philosophy of art, cognitive science, and psychology of art. Lev contributes perspectives from media theory, digital culture studies, digital humanities and computer science, as well as his four decades of experience as an artist creating digital media, including recent work with generative AI. This is the first time all these different perspectives have been combined to analyze cultural AI.
We started working on the book in the summer of 2019, exchanging many messages, commenting on each other's ideas, and sharing section drafts. While each chapter is written by a single author, it incorporates our ongoing discussions.
Given the rapid evolution of generative AI and its wide-reaching impact on art and culture, we made the decision to release each chapter online as soon as it was finished. The first chapter was released in December 2021, and the last one in September 2024.
After all chapters were completed, we did further editing and now releasing the final version as a single PDF.
-- Lev Manovich and Emanuele Arielli, November 2021 – January 2024.
****
All text is available under Attribution-NonCommercial-NoDerivatives 4.0International Creative Commons license: /licenses/by-nc-nd/4.0/
Publication design: Lev Manovich.
Editorial and Production Assistant: Ekaterina Maslova. #GenerativeAI #DigitalCulture #Aesthetics #AIArt
If you are teaching any university classes right now - can you help me? I am collecting global responses for a simple (but large-scale) survey about people's lifestyles & cultural likes. It takes < 10 min to fill out the form. If you can send the link to your students and encourage them to fill it out, this will be great. And of course, feel free to fill it yourself!
The link to the form is in my header.
The aggregated results will be used in a new cultural analytics & visualization project I am working on — and of course, all results and visualizations will be published online.
Lab's earlier projects: /p/projects.html
***
Image in this post - from our project Phototrails (2013) - first to visualize large collections of Instagram photos:
/instagram-cities/
I'm enjoying playing with MoMA exhibitions data - making little dashboards. This one shows the 12 most exhibited artists in MoMA between 1929 and 1990 - divided into 3 periods.
Join me for an online seminar " Medium That Thinks: Generative AI, Media Cognition, and Artistic Creation " on May 13th, 10:00 UTC/GMT (The Department of Film, Theatre & Television, The University of Reading, UK).
Please RSVP to receive link. Registration i go is in my header.
ImagePlot 2.2 in action:
--
1) Mondrian's evolution from his earlier realistic paintings towards his mature style (visualized over time - 1905 to 1918 - in 77 features space using t-SNE);
--
2) Comparing van Gogh's Paris and Arles paintings in 2d feature space - each axis uses one feature.
---
#digitalhumanities #digitalarthistory #culturalanalytics #imageplot
--
Code (since HTML file) and these datasets:
/drive/folders/1F-IShu5Uu_l0xY1wZTKfRc3keJJNtWQV?usp=sharing
I used the same prompt in Midjourney to generate 144 drawings of flowers in a bottle. Here you see the composite (many such images superimposed) - followed by examples of a few individual outputs from.
If you like this experiment, what other subjects should I try that would work well in this format?
I made this dashboard today to keep track of my professional invitations (lectures, texts, exhibitions - from 2024 to now) / - and to try to do dashboard design (my first) with AI. Started design in Perplexity and then continued developing it in Claude.
***
If I knew what I wanted right away, it would have taken about 30-60 minutes - but since I kept experimenting and exploring, it took closer to 3 hours. Usually, in my experience, AI assistants can make a standard design after a first prompt, but if you want to make everything right and/or develop a more sophisticated design, this takes longer.
2D design was always really challenging for me, so being able to work with AI tools to do it better now is very satisfying. And there is something else. When I use AI to generate figurative images from a prompt, the fact that you don't know what you will get is both exciting and frustrating. But design using AI coding works differently. You let AI invent the design, taking advantage of its generative talents. And since it is in code (HTML, JS, React, etc.), you still have 100% "editabiity."
