Speech to text transcription for Mac?
April 15, 2019 8:52 AM   Subscribe

So the industry standard for dictation to text, Nuance's Dragon software, has been discontinued on the Mac platform. There is a an app for iOS, Dragon Anywhere, but it is...lacking. The biggest problem is that it cuts out and stops recording if it drops the internet connection, but doesn't alert you to this fact. So if, like me, you close your eyes or zone out to get in the flow, you can lose thousands of words. My current work around is to record voice memos, then play those back over a speaker into my freaking phone, which then transcribes them. Which is ridiculous. Help?

I don't mind recording voice memos -- I actually like having a back up, as voice transcription software is not the most accurate thing in the world, and I won't always remember a particular turn of phrase when confronted with the Dragon-generated gibberish in the transcribed document -- but having to play them, in real time, into my phone, is, in addition to being silly, double the time commitment. Ideally I would like a service where you can upload audio files and some sort of computer magic transcribes it in ...computer time... but I literally do not know if this is technically possible, let alone available.

Does this exist?
Does this exist for Mac?
If it doesn't exist, is there something less silly that does and that works reasonably well for fiction?

I can continue with my ridiculous work around, but would prefer something less dumb.

Sidenote: Dragon appears to hate Mac users? And so does Apple? Because WOW does this make Macs super inaccessible for lots of people with disabilities. That is just...wow. Really bad. Which is I think why I'm so gobsmacked -- there's GOT to be something better, right? Right?
posted by schadenfrau to Computers & Internet (11 answers total) 8 users marked this as a favorite
Not totally answering your question but just checking because you didn't mention it, did you know that Mac has built in speech to text capability that doesn't rely on an internet connection? I feel like this is why all the aftermarket stuff kinda sucks, because the built in capabilities are pretty good. I've used this when I've had shoulder injuries that required no typing and I've found it pretty good. You might want to make a backup audio track just in case.
posted by jessamyn at 8:57 AM on April 15, 2019 [5 favorites]

Ideally I would like a service where you can upload audio files and some sort of computer magic transcribes it in ...computer time... but I literally do not know if this is technically possible, let alone available.

If using Google Docs is not something you're averse to using, there are third-party Chrome extensions like this one that take advantage of Google Docs' built-in voice typing capability for audio file transcription (AFAIK, voice typing doesn't work offline, unless that's changed recently).

As jessamyn mentions, speech-to-text is already built in to the Mac OS.

See also: Transcribe Audio with Appleā€™s Dictation Tools
posted by mandolin conspiracy at 9:11 AM on April 15, 2019 [3 favorites]

Temi and Otter both do an excellent job of auto-transcribing audio files very, very quickly and generally accurately.
posted by pinochiette at 9:11 AM on April 15, 2019 [2 favorites]

I use youtube for this purpose. I create a blank video with an audio recording, then upload it as unlisted on youtube, and youtube will automatically create captions which you can download.

If your audio is clean the captions are incredibly accurate.
posted by gregr at 9:13 AM on April 15, 2019 [2 favorites]

P.S. Otter also does a large amount of auto-transcription for free each month before you have to start paying (though Temi is also really cheap).
posted by pinochiette at 9:16 AM on April 15, 2019 [2 favorites]

(Jessamyn, do you still have that trick I sent you last year for making YouTube generate captions/transcriptions, in a second language? BABEL FISH, YO)
posted by wenestvedt at 9:21 AM on April 15, 2019 [3 favorites]

Another vote for Otter. It's excellent. I use it on my Android phone for live recording+transcription, and also via its web interface when I need to upload a pre-recorded file. It learns to identify people's voices though I haven't tried that. Free for up to 6 hours per month, pay for more.
posted by anadem at 10:38 AM on April 15, 2019

Okay, I feel the need to talk a bit about Otter. It's fine, but it's not magical.

I've used it quite a bit (last month I hit the limit of the free service). Here's what I've found:

--It likes stereotypical male voices (deeper ones) better than higher-pitched women's voices.

--It likes a flat American accent as opposed to anything else (say a French Canadian accent).

--It tends to repeat words (although this has gotten better) that the speaker didn't repeat.

--It makes some wild, very American assumptions about words it doesn't know, and it doesn't seem to have any ability to use context to figure out words even if the word was used in the previous sentence. (For example, I don't know how many times "barrister" was used in one recording, but despite that, Otter turned it into an assortment of other words and phrases, including "Bear Sterns.")

I still have to go through each transcription, word by word, and it requires me to spend a lot of time fixing mistakes. Note, I do this using actual transcription software and Word because the Otter online interface it not very good--at least I find it clumsy and awkward to use.

Now, if it is just one person using Otter, dictation-style, it may be better, especially if it has a chance to "learn" the user's voice. That's not how I need to use it, so I can't guarantee that assumption.

I haven't used Temi as much. I think it's slightly better than Otter, but not enough that I'm willing to pay to use it (although I have paid in the past).
posted by sardonyx at 11:02 AM on April 15, 2019 [3 favorites]

To the point above, while it's definitely true that Otter makes mistakes (and some of those mistakes are ridiculous), I've been surprised at how well it transcribes various accents and even calls that have terrible connections. I also use it regularly (generally multiple times a day) and at least half the time with people outside the U.S. who don't speak English with an American accent. It does a good job with Indian accents, Chinese accents, German accents, Australian accents, etc. I haven't noticed any difference between male and female speakers.
posted by pinochiette at 1:21 PM on April 15, 2019 [1 favorite]

Win7 Dragon user here, but anyway ... can't you continue using an existing (ie older) Dragon version on your Mac? Assuming the answer is 'no', I use a recorder profile, dictate into a digital recorder (dictaphone), upload the recorder file, and generate the text from the imported file.

I am surprised that you are getting gibberish, that never happens with me, errors yes, but nothing more. Technique? (Dictated, one error - transliteration)
posted by GeeEmm at 3:06 PM on April 15, 2019

As a couple here have pointed out, Mac OS now has an integrated dictation feature. This, I suspect, is why in part Dragon discontinued Dragon for MAC. I use the Mac OS dictation feature often, it essentially uses the same command structure Dragon uses and is robust. You can use it to dictate to multiple programs, Web sites, etc. as well. I'm using it at this very moment to dictate this response, for example. Check it out.
posted by WinstonJulia at 12:19 AM on April 17, 2019 [1 favorite]

« Older Time for another round of Name That Scat   |   Activities for child with limited range of motion Newer »
This thread is closed to new comments.