The art of image processing: imgproxy for ARTE television

ARTE is a unique television channel that started broadcasting in 1992. Every program, documentary, and film it hosts has a cultural slant: about 56% of the programs are documentaries, 19% feature films, dramas, and series, 14% are news-related programs, and 5% feature music and other performing arts. ARTE has also co-produced award-winning films, TV dramas, other programs, concerts, and shows. The programs are broadcast in 6 languages — German, French, Italian, Spanish, Polish, and English and it supports subtitles in these languages.

ARTE considers their arte.tv platform and their social media presence as critical for informing and educating their audience, introducing interactive formats and immersive experiences that speak to a young audience. That’s why they work intensively with several social media channels and post dozens of messages daily. A screenshot supports each post to attract their audience’s attention. In addition to social media images, they need to process pictures, photos, and screenshots on their website and other resources. That is around 500 000 images a day, stored in a MinIO (Amazon S3-compatible storage).

To process these, they initially opted for Thumbor. They were happy with that tool for several years, but they had to admit the fact that the community behind Thumbor was losing their active members fast, and the Thumbor migration to Python 3 was too slow.

Luckily, the ARTE engineering team found imgproxy and started using it. They used the open source version in parallel with Thumbor for a couple of months. (It’s worth noting that some features that imgproxy now supports, like smart cropping, at that moment were already implemented in their previous solution.)

But later, their image processing flow would further necessitate one vital requirement — watermarking. ARTE migrated to imgproxy Pro and made it the primary and sole image solution in their infrastructure. They deployed imgproxy to an internal k8s cluster (not in the cloud to keep their full control), and the charts we provided helped them to deploy the solution relatively easily. The engineering team also found NewRelic monitoring a nice feature to gather metrics from imgproxy.

Today, ARTE’s primary image processing needs are resizing and watermarking. imgproxy resizes images for each client’s frontend (web, mobile applications, smart TV applications) and helps generate images with a blurred ARTE logo for social networks.

Sometimes, they also add program titles onto the images.

Thanks to imgproxy, the ARTE team built a robust image processing system to serve their needs. You can get a solution for your needs, too! Please drop us a line if you want us to help establish these processes in your project.