Atlassian used Transcribe 2.0 in @Loom and found it more accurate at capturing user instructions. Dictate your change requests using Loom, then directly export to Cursor to code it up.
Transcribe 2.0 is available today in the Grok Voice API at $0.10/hr batch, $0.20/hr streaming. Read the blog to learn more: x.ai/news/grok-voice-transcr…
those slices are all clean-ish speech. koenecke ran five major asr systems on matched interview audio and got 0.35 wer for black speakers against 0.19 for white, and each was a leaderboard system at the time. does 2.0 report per dialect wer or only aggregate?
I wish Grok VT had the able to learn my unique sound patterns. I have trouble with R, L, W, and vowels. And there might be a few other sounds.
Though today, I have no need of Grok VT2, voice mode is limited for me, if it can't understand me.
(Though, I might need take a speaking class.)
hyperfunctionality gets in the way of actual functionality - all i really need is Grok 4.2 with persistent memory. You are pitching the product in the wrong way to the wrong people - and when u gave Grok a lobotomy i had to switch to chatgpt as my workhorse - kind regards
Used to bike and chat with Grok but he can hardly grab something. Now I still bike and talk to him, without any problem, just like we talk in a quiet room
I'm really having a hard time with Brock voice chat today. I keep getting the red disconnected chip in mid conversation. It's been happening for about 8 hours now
Used voice mode with Grok Bot last night and the speech recognition was spot on. Felt completely natural. Actually a joy to use. Got more done talking than I do typing. It is truly amazing how good it is.