After getting 90-95% accuracy with Dragon 9.5 preferred and Parallels 3.0 (with MacBoook Pro C2 duo, 10.4.9, 2.4ghz, 4 MB ram), my system has hit a snag. About once a minute the recognition accuracy falls apart for no apparent reason - lasts a minute or so - then goes back to high accuracy - then falls apart again. It's definitely a system problem, there's no ambient noise change or anything like that. This has only occurred since installing the new build 5584 (the one Parallels reissued on 16 January though I don't know if there is any connection). I know there are a few dragon users here - does anyone else have this problem? It is making my standard work dictation system completely unworkable. Tom
Haven't noticed this problem with DNS 9.5 on parallels. I am getting good accuracy continuously. My problem is that parallels goes to 100% CPU and then requires rebooting the mac. This event is routine now, after running for a few hours.
Similar observation seems to implicate build 5584 I have noticed similar problems and odd behavior. In the past, Dragon Naturally Speaking Medical 9.1 SP1 was working acceptably with >95% accuracy. Since installing build 5584, I have noticed the following: - slowed performance overall, accuracy is still there but recognition takes nearly twice as long - intermittent high CPU usage usually at start of dictation but wanes and waxes - segments of the dictation seems to drop out, both when transcribing .wma files and when dictating directly. Would like to know if anyone else has observed this. Since this patch was meant to correct a cut and paste issue after the latest Mac Security updates, I was wondering if the problem was related to the fact that my speech profiles were stored in the Mac Documents location, outside the virtual machine .hdd file. Planning on moving the speech files to be contained within the VM file, but was wondering if others are experiencing this. MacBook Pro, OS X 10.4.11 2.16 GHz C2D, 2GB DNS9 Medical, Plantronics CS50
Seems to have been resolved I had to make a few changes so the answer is not quite clear unfortunately, but my problem seems to have gone away. After moving the speech files to the virtual disk file, recognition was much faster but I still had occasional high CPU usage. I was running out of free disk space (500 MB to 1.5 GB) on the mac side, so I transferred more of "My Documents" folders stored on the mac side, back to the virtual hdd file (which had 20+ GB free space). The transfer was painful because of some files would not transfer because of file name incompatibility and the transfer would job would just stop. Having 3 GB of space and the speech files in the hdd file seems to have fixed the problem. I just tested a 109 word passage from one of my notes and had only 4 errors, with very good responsiveness.
I've managed to solve this problem by trawling through old posts dealing with the best methods of setting up Dragon NaturallySpeaking on parallels. The error I have been committing is simple to rectify and the improvement has been fantastic, so let me share it in case anybody else encounters the same problem. The trick is not to attempt to connect your USB microphone to the virtual machine, but to connect the USB mike to the Macintosh by selecting it in system preferences> sound, which will make it the default microphone for the Mac - and for the virtual machine, as long as you don't attempt to connect it to the Virtual machine. Below is the procedure you should follow, which I have adapted very slightly from a post by webair in the parallels desktop forum on 22 June 2007, which I gratefully acknowledge: 1. In the Mac OS please go to the "System Preferences" - "Sound" - "Input" - highlight you device in the list (this will make it the default audio input device for Mac). 2. Set the input volume in the Preference pane so that the microphone input level peaks near the middle of the bar range. 3. Start Parallels Desktop, choose your VM. In Configuration Editor turn "Autoconnect USB devices" to OFF. [NOTE: in parallels desktop 3.0 Build 5584, you don't seem to get this option; you should choose "connect to the Mac" or "ask me what to do", but if you choose the latter option, whatever you do don't connect the USB mike to the Virtual Machine]. 4. In Configuration Editor, in the Audio tab, choose the Default Audio as the input device. 5. Run the VM. 6. Voice settings for Playback and Recording in the Windows XP Control Panel should be left at Intel[r] Integrated Audio. [I would double check that your USB mike is not selected in the USB options in the devices menu.] If you have already set up a user in NaturallySpeaking, when you try to open it under this new configurations, it will probably ask you to redo your audio settings, which will simply involve doing the volume check and the signal-to-noise check again; this will take a minute or so. The result: the bizarre CPU fluctuations that were destroying my dictation accuracy have vanished. Instead of CPU usage hovering around 100% when using NaturallySpeaking, and then inexplicably crashing, CPU usage runs between about 25% and 50%, the Mac runs nice and cool and dictation accuracy is better than it has ever been. Tom
I am having the same (and worst) problems. I can't get Dragon to work at all. At themost it will work for a couple of minutes. Then it does not recognize speech at all. CPU goes up to 100% It worked fine before installing Leopard...
Hi SergioG The problem seems to have arisen only since you installed a boot camp partition with parallels. Working backwards, this problem you describe (100% cpu usage, distortion, accuracy falling apart) comes from trying to connect the USB mike to the VM. The problem is probably not with Leopard - I suspect it is with the link to the bootcamp partition. Are you somehow connecting the USB headset to the VM through boot camp? Tom
Hi Tom, Thanks for the reply. No, I don't think I am connecting the USB headset via BootCamp... I followed the standard procedure - USB headset selected as default on Mac side, on VM auto connect USB devices set to NO, sound set to Default. Copying the user files to the local C drive on the VM has improved the situation minimally. DNS seemed to work for a few minutes and then went just back to useless behavior. Sergio
Sorry I've run out of ideas. I assume you have also checked that you haven't connected the USB mike in the Devices>USB menu? I can't understand why Nuance or Parallels don't deal with these issues. There's obviously a significant number of users out there who like to run NS on their macs without having to use BootCamp. I guess Nuance doesn't care, with the new MacSpeech coming out that uses the Dragon engine. Tom
Yes, you are correct. Under devices (VM) default audio is selected, not the USB headset... There has to be a way to get this working... Also, under this configuration you select the USB headset or in-line audio for the user configuration in DNS? (I tried both, with the same bad results) sergio
Sergio, I mean the Parallels menu bar at the top of the apple desktop, not the Parallels configuration editor. Select devices>usb and make sure your Plantronics headset is not checked. In NS, USB not inline. Tom
