@simon That's the way to do it. Both with image descriptions and with automatic speech-to-text, editing the machine version is key.