Voice Chat with Local AI

TL;DR: Voice chat in LM Mini lets you talk to local AI out loud. Speech-to-text turns what you say into the prompt, text-to-speech reads the reply, including Kokoro. Continuous conversation keeps going.
So far all of this has been typing, and my thumbs would like a word, so this one is about talking to it out loud.
Voice chat has come in handy many times. When I am cooking, at the park with the kids and my eyes have to stay on them, or I have a thought that will disappear and can’t stop mid walk to spell it. I’ve tried to make it as convenient as possible, like a phone call that you can still function on and are aware of your surroundings.
I recommend an earpiece, just one is enough. Speakerphone in a café is still in a café! I can hear the reply, and with continuous conversation, the AI listens again after it finishes talking, I don’t have to hit the mic every turn.
I can also dictate in one language and hear the reply in another. LM Mini does the speech recognition itself. Want something other than the standard voice packs? Check out Kokoro voices with LM Mini, or ElevenLabs if you are on Pro. Who said options are a bad thing?!
Ready now? Give a voice to your Persona! A tutor and a messy brainstormer shouldn’t sound the same. Unless they’re the same Persona? Then I have questions as to your learning methods. Are you getting the picture now? Private AI on hardware you own, the option to take heavy hitting models on the go with Connect. Give your models their own flair and personality with personas, and now you can speak to them, continuously, like a normal conversation?! Yeah, me too.
Next post is which models actually run well on your phone. Until then, build a persona, give it a voice, maybe put a few in a continuous chat, and then apologize to them as you get interrupted by your kids as you would in a real phone call.
Get LM Mini: lmmini.com