#192: Speech-to-images

I frequently give talks on a variety of subjects (including entrepreneurship, creativity, robotic vision…) It always takes an extra effort to coordinate some images with the material I’m trying to put across via Powerpoint presentations. What is sometimes lacking, though, is an element of extra spontaneity.

Today’s invention is a tool to make talks more attention-grabbing and possibly memorable.


As the presenter speaks, his words are analysed by a computer undertaking rapid speech-to-text analysis. It then performs a realtime image search and displays on the screen one image from, say, the first page of results. With suitable filtering in operation, the scope for embarrassment would be limited and the process might even result in some interesting feedback -if the presenter were to watch the images and comment on them too.

For many talks, it wouldn’t matter if the translation wasn’t that accurate or the recognition rate was low -just as long as images appeared broadly in synch with the main words. This might have the added benefit that speakers would be encouraged to enunciate more clearly (try Googling ‘eh’ and ‘um’).

Finally, this technique also represents a way to add some extra interest to radio broadcasts -it’s a kind of pictorial stream of consciousness rather like the one which visual thinkers naturally adopt.

Comments are closed.