Plotly Studio experiment 2

Blogs in this series are based on the early access preview version of Plotly Studio. Plotly Studio is an AI-powered desktop application by Plotly designed to automate the creation of professional data apps and visualizations.

After my first dive into Plotly Studio, I was curious to push things further — to see how much more control I could gain over the output and how far I could stretch the default behavior. This second experiment is all about that exploration.

The Dataset: MoMA Photography Collection

For this project, I used the photography subset of the MoMA artworks dataset, sourced via Maven Analytics. It’s a familiar dataset for me — my go-to, in fact — because of its richness and versatility. No matter how often I revisit it, I always find new angles to explore.

First Output: Six Visuals

Plotly Studio initially generated six visualizations based on my input. While they offered a decent overview, there was some overlap and not quite the structure I had in mind. I ended up selecting two visuals to keep in what I like to call the “content area”:

  • A bar chart showing medium over time
  • A horizontal bar chart displaying artists and their number of works

These gave me a solid starting point, but I wanted more control over the layout and the visual narrative.

Creating Custom Charts

I built the remaining charts mostly by outlining what I wanted — Plotly Studio handled this surprisingly fast. For the scatterplot in particular, I provided a very specific (almost technical) prompt, which gave me just the structure I was aiming for.

A nice surprise was the card with an artwork image. I wasn’t sure what would happen, but I tried incorporating it anyway — and it worked. One of those happy accidents that makes experimenting worthwhile.

Refining the Output

The two biggest improvements in this round were:

Adjusting the layout, which turned out to be a bit of a puzzle. The order in which you apply changes matters more than expected, especially when you’re working with multiple layers and chart types. But once I figured that out, the structure came together nicely.

Removing filters that were only relevant to a single visual, which simplified the interface and made the story flow better.