In 2012 I was chief digital officer of 24 Ore publishing group. I launched an initiative called 24labs to test new technologies and ideas.
We run an experiment to automatically generate audio/video news.
The idea went as follows: starting from an online search on free news sites to identify topics correlating articles; pass those articles to a summary generator (we were using a canadian software that was designed for spanish, adapted to italian, given its similar sentences’ structure); pass the text to Loquendo’s speech synthesys ad enrich with an avatar using as background thumbnails of graphcs found in those same articles.
At the time, the avatar generation looked quirky, so we never published it, but the generated audio flash news were regularly published to a public web site.
At the time I downloaded a couple of these audios. This is one of them
When we presented these audios to the editorial staff, their comment was that the audio sounded metallic.
At the time they didn’t grasp that the whole process was without human intervention…