Screenshot shows a graph which documents the listening figures (ups and downs) of a Twitter Spaces program. The X axis indicates the number of listeners, the y axis indicates the time of day.
Best Practice

Watch: Twitter Spaces Analytics with FFmpeg, Tesseract, Python

One of the benefits of working with DW ReCO is that the team is so resourceful. When running into a technical challenge, there’s no need to wait for a standard solution to appear in an app store; our colleagues just open their laptops and start hacking. Like Andy Giefer, who was recently looking for a smart way to analyze sessions on Twitter Spaces.

Twitter Spaces is Twitter's answer to Clubhouse, i.e. another important social audio platform–and therefore interesting to DW, as the broadcaster is aiming to expand and diversify its channels.

In the course of their Twitter Spaces research (which included coops with DW's Indonesian and Chinese service), Andy and fellow lab coordinator Daniela Späth eventually connected with other media innovators, namely Charlotte Voss and Moritz Metz at DLF lab (hi there and thanks again!). They provided us with all kinds of useful insights and eventually invited Andy to observe, record, and analyze the Twitter Space they had scheduled on the night of the German Federal elections.

And here's what all of us wanted to find out with the help of this session:

  • Is there a way to track the number of people tuning into the conversation over time? 
  • How long does the Space need to run to reach a stable number of listeners?
  • Can we build a nice graph to visualize this?
  • Can we also use the graph to identify weak moments that made people lose interest in the conversation?

Watch the video below to find out how Andy found answers to these questions with the help of FFmpeg, Tesseract, Python scripts, and–of course–a proper geek mindset.

In case you want to dig deeper and create graphs yourself, you can also find all necessary code and instructions on Github.

Logo Deutsche Welle
DW Innovation