sul-dlss / vt-arclight

An Arclight-based discovery application for materials from the Virtual Tribunals project

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Create and present audio transcripts for WCAG 2.1 compliance

jcoyne opened this issue · comments

Currently the audio files do not have transcripts which are required by the Web Content Accessibility Guidelines (WCAG). https://www.w3.org/WAI/WCAG21/Understanding/captions-prerecorded.html

There may be some related policy that applies. See https://uit.stanford.edu/accessibility/policy

Since there are courtroom transcripts that mirror the recordings, we're putting this on the back-burner for now. We have this listed as a "must have" in our features priorities doc for the future and will get back to it during a future workcycle.

@pleonard212 has done some work testing out OpenAI's Whisper on SDR content with fairly good results; noting here since we might be able to leverage that strategy later on when this comes back up.

I have the large-v2 model running on my office machine (uses ~ 12gigs of VRAM) and am happy to run some tests if you'd like!