openhuman v0.54.0 is here and it's one of the biggest things we've shipped

today we pushed openhuman v0.54.0. 🚀

this is the biggest release we've shipped. 101+ bug fixes, major new capabilities, and the kind of stability and security work that makes every future release safer. let's get into it.

fully offline voice. your data stays on your device. - voice is now completely local. speech to text via whisper, text to speech via piper. no cloud, no api calls, no data leaving your machine.

*we also shipped a podcast generation toolkit * - the agent can literally talk to you, not just text at you. and you can now customize tiny's voice and color, which is a small thing that somehow makes the whole thing feel alive.

11 languages. global by default.
tiny can now speak and respond in 11 languages including mandarin, arabic, spanish, hindi, italian, french, and more. this isn't just ui translation — the agent actually converses in these languages.

we've been wanting to do this for a while. ai assistants that only work well in english are leaving a huge chunk of the world behind.

the agent can manage its own todos
this one is a big deal for deep work use cases.

the agent now has a full todo board with crud — it can track tasks, orchestrate multi-step workflows, and run things in the background while you focus. we've been calling this deep work automation internally and it's the foundation for a lot of what we're building toward.

stability, security, and actually visible errors
the app is now meaningfully more stable across windows, macos, and linux.

on the security side: path traversal prevention, dns-aware url validation, windows secret key self-repair. we also cut a lot of noisy sentry errors — the kind of thing that drowns out the real signal. now when something breaks, we actually see it.

testing coverage is substantially stronger. this makes future releases safer and faster to ship.

dark mode. light mode. finally.
yes, it's in this release too.

what's next
this release reflects what we believe building toward agi actually looks like — not just model benchmarks, but systems that are private by default, persistent, and genuinely useful in daily life.

download v0.54.0 here →

a huge thank you to the devs, the community, the contributors, and honestly myself for the all-nighters. this one was worth it. 🫶