Want Better Speech Input? Speed is Key

I periodically get questions about the best computer setup for speech input.

And I always advocate going way above the minimum specs for speech, especially if you’re planning on using speech input beyond simply putting words on the screen. Having a fast computer that interprets what you’re saying quickly makes for a much better experience – especially if you are using the computer for command-and-control: navigating, selecting, formatting and correcting text, and controlling Windows elements including windows, menus and buttons.

Here’s how speed matters. Picture walking across the room and having to pause slightly between each step. It may be tolerable for a while, but long-term it’s frustrating, right? Keyboard and mouse input is usually instantaneous. Speech input, not so much. Even a slight pause while the software recognizes and executes a speech command gets frustrating over time. The faster the computer, the shorter the pause.

This is because it takes a lot of processing power for the speech recognition program to process the sounds you make, and you use speech input alongside anything else you’re doing on your computer. It takes a relatively fast computer to run speech input and your other programs well.

Speech input is prone to mistakes – as the software gets better the accuracy rate goes up, but it’s not 100 percent, and isn’t likely to be for some time. Being misheard happens periodically with speech input – just as it does when you’re talking to people and usually more often – and it’s something you have to watch for. A faster computer means more accurate speech input. Dragon has a speed versus accuracy slider bar in Options that lets you boost one or the other. Ideally, you want as much of both as possible. If you use add-ons to Dragon or add a lot of your own macros they’ll take some extra processing power.

With that in mind, you need a relatively fast computer for a really good speech input experience. Here are the numbers that count:

Processor speed matters, and the size of the L3 cache on the processor matters the most. Memory size and disk speed also matter.

If space and portability don’t matter to you and price does, keep in mind that you can get more desktop power for the price.

Looking at today’s desktops, here’s an ideal desktop set up. You can get a desktop with these specs for around $1200.

Key specs for a desktop computer that will run speech recognition very well:
1. An i7 processor with a 9, or better yet, 12 megabyte cache
2. 32 gigabytes of memory
Memory comes at different speeds and in different size chunks. The faster the better. DDR4 2666MHz is a good choice. The larger the chunk the more potential for upgrading, for instance 32 gigabytes as 16GB + 16GB chunks is better than 32 gigabytes as 8GB + 8GB + 8GB + 8GB.
3. 1 terabyte, 7200-speed Hard Drive
Hard drives are considerably cheaper than solid-state drives and are probably more reliable long-term. To keep your hard drive running well, defragment regularly and don’t fill it more than 80-90 percent.
4. The Pro version of Windows

For many years I had to have a heavy laptop to get enough processing speed to run speech even tolerably well. But in the last few years processor speed has been fast enough on light laptops to run speech input well. Laptops are lower power than desktops, so you won’t be able to get as powerful a processor.

The key to a light laptop is a solid-state drive. A solid-state drive, which is more expensive than a hard disk, especially because it needs to be kept less than half full. The bottom line is a light laptop that runs speech recognition well will cost about $2,100, or about $1,000 more than a desktop that runs it very well.

And it’s also important to pay attention to battery settings on a laptop. Sometimes battery saving modes make the computer, and thus speech, run slower. Plug in when you can, change your settings so the processor runs at full speed on battery, and recognize that your battery won’t last as long as it does in power saving mode.

Key specs for a light laptop that will run speech recognition well:
1. An i7 Processor with at least a 4 megabyte cache  – the more the better
2. At least 16 gigabytes of memory – the more the better
3. A solid-state disk that’s twice as big as you need.
A solid-state disk is lighter and also faster than a regular 7200 RPM hard drive, but only if you keep it less than half full. So when you buy a solid-state disk, remember that the usable space on that disk is less than half. If you need more storage, use a removable hard disk.
4. The Pro version of Windows