A comprehensive answer to this question is in the works. Our documentation team is just coming together. This is my personal take, but others might put it differently. At the bottom, I’ll link to a feature chart at our technical docs, which describes them thusly:
OpenVoiceOS ready to use images come in several flavours; The buildroot version, being the minimal consumer type of image and the Manjaro version, being the full distribution easy / easier for developing. a headless raspbian image is also maintained by the community
The Buildroot-based image is the original OpenVoiceOS from way back at the top of this thread. It is
- Pi-first, including the Mark II (several OVOS devs were testing on the Mk2 SDKv1 for a long time, and some still are)
- very much an appliance
The Manjaro-based image is a full-featured Manjaro-ARM spin, also configured to run on a Pi-based smart speaker, and also supporting the Mark II. It’s quite a bit heavier than the Buildroot image, but, being a full OS, it’s the easier place to start tinkering.
Finally, the Buildroot-based system takes a long time to build. For that reason, the image linked on our site is currently outdated. Our most recent fundraising target was a budget for a buildserver, and was successful, but we’re still in the process of incorporating. Once that happens, we expect to go from extremely infrequent updates to extremely frequent.
We’re too young to have developed an LTS policy, but we have no plans to retire either system. Rather, once OVOS is incorporated, and we’re finished with the current round of expansion, our capacity for user support will be greatly improved. Different members of the core team are principally responsible for each, and the tricky work is usually about the hardware rather than the operating system, so we haven’t found much difficulty in maintaining the two.
I realize this doesn’t actually offer much in the way of guidance, and I’m sorry for that. Both systems are meant to support the whole OVOS stack, so it’s tough to say, “Use that one.” Here’s that feature chart, covering our speaker images, a couple of our downstream friends’ images, and MycroftAI’s.