Have you tried Soniox for speech recognition? It supports Croatian. Or are you just looking for self-hosted open-source models? Soniox is very cheap ($0.1/h for async, $0.12/h for real-time) and you get $200 free credits on signup.
I meant in general purpose tools from Google and Apple. Most of this assistant and "AI" stuff is practically useless for me because I refuse to talk to my devices in English.
In Android Auto / CarPlay I can't even get voice guidance that works properly, much less reading notifications, or composing a reply using STT
https://soniox.com/
Disclaimer: I used to work for Soniox