@thomasfuchs this is something I would argue should be provided by the OS for all images / media. Feels like a great use of ML to improve accessibility for all similar cases.