Improvisational Powerpoint, or Speech-To-Pictogram
Uses Web Speech Recognition API to create a generated-on-the-fly presentation. Talk about whatever you want, and the presentation will follow you!
- Creates a closed-captioning-style transcript of the words spoken
- A Natural Language Processing library plucks out interesting topics or nouns
- A background image is created from those nouns via a call to the Flickr API
Requires desktop Chrome. If you decide to try it for realsies, make sure you're close to the mic (a lavalier microphone works best) and speak clearly and not too fast.
yarn but only to keep it up to date, there's no node involved otherwise.
This must be run on a SSL-enabled web server (e.g., using HTTPS and not directly from a file); those are Chrome's restrictions.
Unpack to the desired directory, then
npm install to load the dependencies.
This was a one-off prototype and isn't under active development. Still, it would be neat to:
- Build it out as a plug-in for reveal.js
- Talk with accessibility specialists; could something like this be a good-enough solution for deaf audience members when a real interpreter isn't possible? (Talking about the transcription here, not the silly background images.)
- Include credits/links to original Flickr contributors for the generated background images
- SpeechRecognition API occasionally hangs/craps out and it's not clear why