I just got accepted to the Speech-to-Text beta, and tried to use it, but it wouldn't generate a caption track. (The "creating" wheel has been running for about a half an hour as I write this, this is my second attempt after closing and restarting the program.)
It's been failing to generate caption tracks for a couple weeks, actually. I tried rolling back to earlier stages of the beta that I knew it had worked in, but it didn't fix the problem. I've since been just captioning using graphics.
Don't know if this will help you, but I saw the same issue this tonight (I only got the Beta yesterday, as part of the Speech-to-text trial) ... for me, I'd got to the point that the cloud service had returned a transcription, and was also trying to generate the caption track.
First attempt I made, it appeared not to start (spinning 'creating' wheel) - but I switched tabs in the Text box to 'Captions', clicked three buttons in the top right, and clicked 're-transcribe' ... that seemed to work, as I got a new dialog box to determine the caption paremeters, and ultimately a populated caption track.
Second attempt was on a slightly longer film - similarly slow to start, so I did the same as above again, and it worked the second time.