Explore local LLMs for a11y
We currently don't have a good story for text-to-speech (the screen reader voice is terrible), speech-to-text (we have nothing), or translation (we have nothing). A lot of this was very hard to do locally until recently, but things like Live Captions and Firefox translations are examples of using fancy new LLM tech to enable these features completely offline.
It'd be great to explore the space a bit and consider how we could use this to provide system services for, e.g.
- Reading text out loud
- Screen reader
- Narration feature in e.g. web browsers or writing apps
- Transcribing audio
- Live subtitles as a system feature
- Transcribing voice memos
- Could be the basis for a fully speech interaction mode for a11y?
- Translations
- Fancier spell check/grammar check
Edited by Tobias Bernard