Seer – Open-source local AI image descriptions for screen readers (no API key)(github.com)
1 pointby chiachi5 hours ago1 comment
  • chiachi5 hours ago
    Built this because most image description tools require a paid cloud API. Seer runs PaliGemma2 entirely on your machine — no account, no internet after setup.

      Currently Chrome only, ~30s per image on CPU. Looking for feedback from NVDA/JAWS/VoiceOver users especially.