Wikipedia Autovideo

This Java program automatically creates narrated videos starting from Wikipedia pages, using images from the page itself and other sources (Pixabay), and doing speech synthesis for the text using Amazon Polly API. If you want to read an in-depth description of how this tool works you can read more here.

This is an example resulting video, for part of the page of the city of Florence: https://www.youtube.com/watch?v=IROxKUeGi0c On the same channel you can find numerous other examples of such videos.

How to run

You need a Java version >= 8 and Maven installed to build the project (you can also build without Maven manually, but you'll need to take the approriate steps).

1. Run mvn package in the root directory of the project. This should build without errors and create a runnable jar file named autovideo-0.0.1-SNAPSHOT-jar-with-dependencies.jar in the target directory.
2. Before running, you need to rename the autovideo.conf.example file into autovideo.conf and fill in your API keys for Pixabay and Amazon Polly.
3. At this point you are ready to run, for example:

java -jar target/autovideo-0.0.1-SNAPSHOT-jar-with-dependencies.jar New_York neural

Note: run the command from the root directory of the project, as shown, and not form inside the target directory. The code relies on relative paths that wouldn't work otherwise.

This will retrieve the New York Wikipedia page and create the video using Amazon neural engine for speech synthesis. This is higher quality but more expensive; the other option is to use standard instead of neural, especially for debug, as it is cheaper.

Dependencies

This tool uses ffmpeg, so you need it installed with a version that's compatible with the ffmpeg filters that we are using. Can't tell an exact number but any version released after 2020 should be ok.

Also, to perform image conversions (e.g. SVG to JPG) the tool uses the convert command from imagemagick.

Output

The program outputs is stored in the output/final directory. As the program runs, you will see files like video.final.0.mp4, video.final.1.mp4 in the output directory: these are just intermediate steps that are later joined in a single video, but you can still watch them to see how the result is coming out.

Before starting with the actual video generation, the program also creates an HTML file in output/final which provides a preview of the final video. It shows which images will end up in the video, how the video is divided in sections, with an estimated length for each one, and what text will be used.