Fora

WORDT DEEL VAN DE COMMUNITY
REGISTREERINLOGGEN

Model S Software Note

This morning I received an email from Tesla with the same subject. I don't know if it went out to all owners or just folks like me who are still on 4.1 because they didn't want to lose the sleep mode. Anyway, there was a paragraph about a future update that I thought might be of interest to the forum at large:

"Our ability to maintain seamless functionality and introduce new features is contingent on customers upgrading to the latest software. For example, an upcoming change to Google voice recognition technology on April 30, 2013 will disrupt the Voice Command feature in Model S software versions 4.2 and earlier. To ensure the best ownership experience without interruption to features, Tesla recommends you upgrade your software."

So, that answers the question about how far back releases will be supported (and no, I didn't get that email!), but on the plus side, it sounds like Google voice recognition is about to improve (which is good because mine has sucked so far!).

I don't know what the fuss is. I still have a pc running windows 2.0. To make sure it runs properly, i haven't powered it up in 25 years.

Um....Google voice transcribes my phone messages. It's still hit or miss, and that's with people speaking directly into a phone. I hope we're talking a new Google Mobility Solution for Voice or I might be hanging out with Steven.

The difference in voice recognition results between Google Now (on Android) and Google Voice (voice mail transcribing) is night-and-day. Google Now is almost always spot-on, comparable or better than SIRI.

On the other hand, I get a good laugh reading Google Voice transcriptions -- sometimes they're completely out of the ballpark.

I'm hoping Tesla is tapping into the same services supporting Google Now.

Transcription of voice mail is after the voice has been encoded with something like G722 which basically filters you down to 50Hz-7kHz or so, and then compresses that lossily. If you're using a cellphone to speak to the voicemail, then you first got GSM'd, so you actually probably got lossily filtered and compressed twice before the audio recognition had a chance.

By comparison, the speech recognition stuff is recording you at high quality on the device, compressing less aggressively with less loss (no real-time requirement helps), and doing so in a way that is optimized to not lose precisely those bits of signal that voice recognition cares about.

That's why one sucks and the other works pretty well. The quality of the audio being analyzed is vastly different.

Just an FYI. I took in my Model S for minor servicing today for a faulty TPMS sensor. While I was there, they informed me they are going to routinely replace some R and L rear axle nuts. In addition, they are updating the DC/DC Bootloader. Apparently, this is the connection between the main and 12V battery that is giving some owners problems. I will try and find out more tomorrow about exactly what they have done. It seems to me regular scheduled maintenance is a good idea.

Hopefully this is more "all around" voice integration into the system. There are so many opportunities, aside from playing a song or navigating to a place.

"Camera on bottom" .. "Set temp to 68" .. "Turn on parking lights" .. "Set height to very high" .. "Open left garage"

The more we can keep our fingers off the screen while driving, the better.

I'm still waiting for my update (on 4.2) so hopefully this means that the update will show up soon!

Does Tesla maintain older release notes somewhere on there site?


X Deutschland Site Besuchen