Great! Take a look at local models, I just added multiple locally running models to nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyv9kh2uewd9hj7qghwaehxw309aex2mrp0yh8qunfd4skctnwv46z7qg6waehxw309aexjcn09ejh2tnwdaehgunfvyhxzurs9uqzp5daxvenwv7ucsglpm5f8vuts530cr0zylllgkwejpzvak0x2kqmehusjv
Needs more polishing, was up late last night getting it done.
Can do translations, sentiment, speech to text, text to speech, with full privacy, all running locally.
These models are just 60-150MB, works good enough.
Text generation and chat can be done too but require 2-3GB models.