Image-to-text working out of the box

github.com

Oh wow! I try IT later, do i need a API key or how does it Work?

All you need is a linux box :) I use kubuntu.com 24.04.3LTS with updates. it runs locally, and expected run time for 4 AI models and 2 questions (description, keywords) is about 9.5 days for 22K 1080p+ photos. But that's on a box that cost me 4 grand back in 2023. Yeah, just the box. Already had the monitors. I got 64GB RAM and 2TB nvme storage, on this board : https://www.amazon.nl/ASUS-CROSSHAIR-X670E-moederbord-geheug...

You can get started with https://github.com/rene-ajm-veerman/nicerApp-WebOS-5.10.z/bl...

and https://grok.com/share/c2hhcmQtMw_366bdd45-7dbf-47fa-8e0a-fd...

Thx for answer! So easy over virtualbox?

Thats cool, i can really use it for my Work! Thx!

yeah, virtual box should work, but here's a tip : make sure all of your scripts (i recommend asking grok.com for scripts, and asking him again to fine tune it) are (A) resumable and (B) save their overall results often (to a real disk).

Image-to-text takes "forever". In this case, I already let grok make the script that starts this thread, resumable and often-saving.

Reply on news.ycombinator.com