Open Cyclonit opened 4 years ago
Hi,
currently there is no such feature, sorry. It is a very complicated topic, especially when combined with open-source tech. Is there some software you're already using for this?
Maybe there are other ways to solve your problem like pass phrases or pin codes in services. You think this could be an option?
Cu, Florian
Hi Florian,
thanks for the quick reply. I have just started work on my home assistant/automation setup and am comparing various systems as to what serves our needs best.
We make extensive use of to-do lists, calendars and alarms. Having to specify whose lists or calenders we want to access would be tedious and very repetetive.
Is there any project working on this at the moment? Something that needs training to differentiate a couple different people would be fine for us.
Hi,
I've seen a few experimental things over the time, but so far no open-source speaker recognition system that would fit the requirements of SEPIA. I will keep my eyes open though ;-)
Something that could work is face recognition ^^. The open technology for that is much more advanced and even runs on systems like the Raspberry Pi. It would be a larger side project for SEPIA but I think its possible :-)
We make extensive use of to-do lists, calendars and alarms. Having to specify whose lists or calenders we want to access would be tedious and very repetetive.
I see. Depending on your requirements for security I could think of a few options for SEPIA, e.g.:
@Cyclonit There is a new player in town (actually a very old one with a new project ^^) and some interesting developments coming up: https://github.com/alphacep/vosk-api
I will update this topic as soon as I know more ;-)
The new SEPIA STT Server has support for Vosk including speaker recognition. I haven't played around with the feature much so I'm not sure about the performance but maybe there is someone out there interested in doing some experiments ;-)
Hi,
I am considering SEPIA as one option for automating my home, but some use cases require speaker recognition to work smoothly. Is there some hidden support for this in SEPIA already or is it a planned feature?
Kind regards