NatSpeak is finicky, and people who like it best have fast computers with at least 8MB ram, good sound systems, and don't run many other applications while they are using it.
Install winbind. Activate 32-bit registries. Create a 32-bit wineprefix. (NatSpeak "runs on" 64-bit systems, but the underlying program is 32-bit.)
If you do not use pulseaudio, set winetricks sound=alsa.
NatSpeak's built-in word processor (DragonPad) has limited functionality, saves to RTF as a default, and usually crashes, but if you want to try using it do the following:
winecfg - under Libraries, set mscoree to (native, built-in)
You are ready to install.
Installation goes smoothly. When installing I uncheck any options that do not impact functionality, including automatic updates and the like.
Training goes well. I skip training myself, but most users do it, and it works fine.
When I open the program for the first time, a help box and a sidebar open up. I close them permanently.
And I am ready to go!
What does not
Text recognition only inputs directly into other programs running in wine. This is correct functionality, not a bug.
Use NatSpeak's Dictation Box for unsupported applications.
To activate The Dictation Box opens anywhere. And if you check the box "keep text in buffer" you can say "Transfer this" and then use CTL-v to paste text anywhere.
What was not tested
The program has a lot of function that is extraneous to its core mission. I have only tested to find if speech recognition works quickly and accurately, and that the program doesn't crash much.
HINTS FOR USE:
(1) Linux's support of USB sound devices is patchy. Everyone says use a USB sound device, but I have bought a couple of very expensive USB microphones "guaranteed" to give perfect sound with NatSpeak, and find that just using the computer's included sound system works best. If you want a headset, try plugging it in to the microphone jack.
(2) Insiders know that NatSpeak's user profiles degrade after a while. Save yours sparingly. This is true when running on Windows also.
(3) Speak in a normal tone of voice. Don't think you can mutter quietly into any microphone and get good recognition.
|Operating system||Test date||Wine version||Installs?||Runs?||Used|
|Current||Debian GNU/Linux 9.x "Stretch" x86_64||Oct 12 2018||3.0.2||Yes||Yes||Yes||Silver||Susan Cragin|
|Show||Ubuntu 14.10 "Utopic" i386 (+ variants like Kubuntu)||May 23 2014||1.7.19||Yes||Yes||Silver||Susan Cragin|
|Show||Linux Mint 17 "Qiana" x86_64||Jul 29 2014||1.6.2||Yes||No||Garbage||Jay Starkman|
|Show||Ubuntu 13.10 "Saucy" amd64 (+ variants like Kubuntu)||Nov 20 2013||1.7.4||Yes||Yes||Silver||Susan Cragin|
|Show||Ubuntu 13.04 "Raring" amd64 (+ variants like Kubuntu)||Apr 12 2013||1.5.27||Yes||Yes||Silver||Susan Cragin|
As of wine 3.6(staging), and possibly earlier, NatSpeak 12.5 installs and runs under a 64-bit wineprefix.
I installed it on 28-Apr-2018 and the only special concession I made to the wineprefix was to uninstall mono. Installing dotnet 40 with winetricks also works. Both optiions tell the program that an inadequate version of net is installed, and prompt the program to install its own.
After the installation, it says you have to reboot. I kill wineserver at the terminal and then start the installation process over. It does, bypassing the net installation.
All in all, NatSpeak works wonderfully now, and version 12.5 is very accurate and the latest low-end (and cheap) version to work well.
Dragon NaturallySpeaking 12.5 is actually not obsolete, since many people still use it, and it was the last Dragon Naturally Speaking to run well on wine.
Versions 13 and 15, both newer, cannot be installed.
12.5 can be installed. However, it no longer runs. A bug crops up after you have created a user profile, and before you first use it.
So there is no working version of Dragon Naturally Speaking right now.
I last tested 12.5 with wine-4.3 (Staging) and it does not run.