*** A special note for Dragon for Mac Forum Users ***
Nuance is pleased to announce our new Product Forum User Community - the new Dragon Dictate community will replace these existing Dragon for Mac Forums.
To access the new Online Forums, you must first sign up for an account through our Support Portal here. After activating your new account, or if you already have an account, you may use THIS LINK to log into the Community.
Please note that as of June 18, 2012 this Dragon for Mac Forum has been set to read-only in order to retain all historical information and post details. Any new content or updates to existing content will no longer be permitted at this time. Please visit the new community link listed above to post new content and start new topics of discussion.
tspencer
Member
It's getting late where I'm at, but I was ecstatic to finally find a [minorly hacked] solution for doing transcription in Dictate. The results were positively astounding to me, with several different devices and a couple different mics. All goes to show me how good Dictate really is...
No time to post now because I've gotta hit the sack, all I'll say is it involves Soundflower and GarageBand. Full explanation tomorrow :)
Cheers!
tspencer
Member
First of All: Sorry if this was at all suspenseful. By the time I had figured it out, I was too tired to retrace my steps, much less write them down coherently. That being said, it is pretty easy to set up.
***NOTE: This may not be as streamlined a setup as you want, it may not work for everyone. It requires a basic understanding of how to use a Mac. If I have made anything unclear (which is likely the case, mention it to me and I'll clarify. Other than that, go for it.***
Step 1 - Soundflower:
- Download and install Soundflower (from here). This also installs an application called Soundflowerbed. You don't need to run this application, it simply assists in setting inputs and outputs for Soundflower. (What soundflower does is allow you to pass output of application(s) or microphone(s) to input of other audio device(s)). If you have Snow Leopard you can change inputs and outputs by option-clicking the audio menu-bar item.
Step 2 - Set Up GarageBand:
- Open GarageBand. Open GarageBand preferences (GarageBand>Preferences or ⌘-,). Select
the Audio/MIDI tab and set the output to "Soundflower (2ch)." Set the input to the
microphone you normally use with Dictate.
- Add a real instrument track if there isn't one already. Make sure that the track doesn't have
any effects on it (Choosing Vocals>No Effects from the track info window is the easiest way
to verify this).
- The last two things to do in GarageBand are to record-enable the track and to make sure
that "Monitor:" is on. This is set in the track info sidebar. Verify this by talking into your
microphone, and the garageband track should display the audio levels as you speak.
Step 3 - Set Up Dictate:
- Open Dictate. Open the "Profiles..." window. Add a new profile and can call it whatever you
want. The important thing here is to set the "Microphone" to "Soundflower
(2ch)" - no, it's not a microphone, but garageband will make it into a mic for us.
- Proceed with microphone setup and Voice Training (I know, *sigh* what a burden, I actually
have to spend another 5 minutes creating a profile? - it's worth it).
- If you have problems here, it's likely because something's not quite configured right.
I can try to troubleshoot if you have questions as I'm relatively familiar with
GarageBand and Soundflower
Step 4 - Extension and Problems:
- That's right, you're done.
- But it's not all that useful if it's just your normal microphone going through garageband. So
how's transcription involved? Well first, try recording some dictation in GarageBand with
Dictate's "microphone" off or asleep. Now play the recording back with the "microphone"
back on. You probably won't hear it, but Dictate will and if it works anything like my
[low quality] microphone it'll be brilliantly accurate. Cool, but that's not very useful.Touché.
- Quit GarageBand. Option-Click the Audio icon in the menubar. Set the output to
"Soundflower (2ch)". This is the System Output. So ... everything you play, Dictate will listen.
Songs, podcasts, obviously not useful. But recordings copied to the computer? - Yes, please.
I recorded a voice memo on an iPod Touch with the iPhone headphones (even lower quality)
and it was still 100% accurate for the short clip.
- One glaring problem is that if the system out setting is soundflower, you can't hear anything. Two
solutions: Solution One is to reverse the input and output settings you had in GarageBand. Again make sure monitor is on and that the track is record-enabled. The problem here is that GarageBand has to berunning to hear anything. Something of a hassle. Solution Two: Launch soundflowerbed. Flower menubar icon appears. Click on it and right below the Soundflower (2ch) and the purple flower icon, you can choose what is basically a monitor of what soundflower is "hearing." Choose whatever output you want and now you can monitor what the computer is outputting again.
Enjoy Everyone! If you have any questions contact me I'd love to be able to help you work out any kinks in the setup.
tspencer
abledoc
Member
Impressive work, congratulations!
Hectors Dad
Member
Can I clarify - you drop a previously recorded sound file into GarageBand and when you play it Dictate will convert it to text?
Presumably as you have to train it this will only work with sound files that you will have recorded yourself i.e. it has to be your voice?
Wintanclan
Member
Thank you tspencer for this idea for recordings you already have on the computer.
While I still don't make use of this regularly, I have experimented with using transcription of my own voice. My setup is analog and pretty pedestrian, but it works: I first dictate into my iPhone, and for transcription I directly connect the headphone jack of the iPhone with the microphone jack of my "certified" USB headset. so MacSpeech thinks I still use my standard microphone, and transcriptions have been surprisingly accurate...
Martin C
Member
Which brings me to a thought I had last month about headsets.
The bit that Dictate seems to care about is better sound encoding offered by the dedicated chip in the USB dongle. (I assume this does noise cancelation, smoothing of the signal and attenuation)
How the sound is transmitted from headset to your laptop might be secondary.
Has anyone tried a radio mic with 3.5mm receiver output hooked up to the USB encoder? Or even a bluetooth mic transmission to bluetooth receiver plugged into that encoder?
Cheap mic + radio transmit + existing dongle might circumvent needing to spend on an expensive dedicated integrated receiver.
Thanks,
Martin.
Trace
Member
Wait, WINTANCLAN, ya lost me, sorry... you said, "first I dictate into my iPhone, and for transcription I directly connect the headphone jack of the iPhone with the microphone jack of my "certified" USB headset. so MacSpeech thinks I still use my standard microphone, and transcriptions have been surprisingly accurate..." How does one connect the headphone jack of the iPhone to the microphone jack of a USB-headset? What cable are you using? sorry for not getting it right-off! :-) trace
Sentient Paradox
Member
Hey, I'm new to Macspeech. I'm a songwriter and my Mac Pro is set up as a DAW with Pro Tools running through an M-Audio Projectmix. I've got over $400 worth of professional microphones, so the prospect of being relegated to using the cheap mic and headset that came with MSD seems, well, unprofessional.
Hadn't thought of using Soundflower to force MSD to hear the mics.
Too bad I just upgraded to Snow Leopard though. *sniffle!*
Soundflower isn't yet compatible with SL. MAN! I hope the Soundflower team gets on this problem SOON!
jayg
Member
I've got over $400 worth of professional microphones, so the prospect of being relegated to using the cheap mic and headset that came with MSD seems, well, unprofessional.
You can try using a USB adapter for input. That's what MacSpeech Dictate expects.
Donkeyoatay
Member
"Hadn't thought of using Soundflower to force MSD to hear the mics.
Too bad I just upgraded to Snow Leopard though. *sniffle!*
Soundflower isn't yet compatible with SL. MAN! I hope the Soundflower team gets on this problem SOON!"
I am using Soundflower with Snow Leopard with no problems whatsoever.
Sentient Paradox
Member
"I've got over $400 worth of professional microphones, so the prospect of being relegated to using the cheap mic and headset that came with MSD seems, well, unprofessional.
You can try using a USB adapter for input. That's what MacSpeech Dictate expects."
jayg, I got a USB adapter with headset that came with MSD. It works fine.
My point is that it seems silly to be forced to use it when I already have far superior equipment that I spent a small fortune for. At least the headset was free, so I'm not complaining about the expense. I'm not complaining or blaming MSD either. I don't expect you guys to adapt to every system out there.
It's just a bit frustrating when you have $400 worth of microphones, and a $1500 sound board right there, yet must dig out a headset that looks like it came from Radio Shack in order to use MSD.
I am thankful that tspencer made this post to remind me there is a solution. I can't wait until Soundflower fixes the compatibility issue, and I'm sure they will eventually.
Donkeyoatay:
Which Mac model are you using?
I neglected to mention this is a Mac Pro.
Soundflower does not install properly at all since upgrading to SL on my Mac Pro. The problem has been reported to Soundflower by me and several other Mac Pro users, but they have not yet responded.
Donkeyoatay
Member
Donkeyoatay:
Which Mac model are you using?
I neglected to mention this is a Mac Pro.
Soundflower does not install properly at all since upgrading to SL on my Mac Pro. The problem has been reported to Soundflower by me and several other Mac Pro users, but they have not yet responded.
I am using a new 13" MacBook Pro.
Sentient Paradox
Member
Donkeyoatay:
That could very well be the difference. Mine is one of the earliest Mac Pro models that shipped with Leopard (January 2008). This isn't the first conflict I've had because of the particular configuration of this model. Can't remember off hand exactly why. Could have been something to do with the Intel Core-Duo processors. If I remember correctly something changed about those not long after I bought this one.
Fortunately, all other conflicts have been resolved by now. Hopefully Soundflower will follow suit soon.
In the meantime, the headset will have to do.
WJC
Member
Can any of you guys who used this setup in the past please post the settings that you used for the sound buffers. I am trying to use Dragon Dictate 2.0 on a MacPro under Snow Leopard. It doesn't like the audio from my microphone (Philips Speechmike Pro) and I don't want to give up the mike.
Linein has a range of buffer sizes at both the sound in and sound out ends. Soundflower has a similar range of choices at the sound in stage. I have played with them for a bit and they seem to cause a big difference, but there are a huge number of possible combinations. Anyone have any ideas on which is the best?
Archer
Member
Increasing the buffer size increases latency. If you make the buffer too small, then you run the risk of the consumer side of the buffer overrunning and causing problems. If you set the buffer size too large, the latency can get so high that your recognition will run significantly behind what you're saying, and cause more confusion. If I were doing this, and I'm not, I would set one buffer to a fairly small amount, say 4000 samples, and then vary the buffer size in the other program until I was happy with the results.
© 2002-2010 Nuance Communications, Inc. All rights reserved.
