Mazty said:
Yep. There's a good reason voice recognition is based from servers rather then local systems. When you have to account for sentence structure, accent, pitch etc there are a lot of variables, and that's just to understand what has been said. A good example is that Watson, the supercomputer that was used in Jepoardy, didn't actually translate the questions - it had them digitally sent at the same time as when the question was asked. |
Processor requirements for Dragon Naturally Speaking:
CPU: We recommend 2.2 GHz Intel® dual core or equivalent AMD processor. (Minimum 1 GHz Intel® Pentium® or equivalent AMD processor or 1.66 GHz Intel® Atom® processor). NOTE: Faster processors yield faster performance. (IMPORTANT: SSE2 instruction set is required)
Voice Recognition does not require a ton of processing. It's the AI algorithms that happen after the voice recognition to come up with the response that require a bit of processing. I'm sure a game could utilize cloud-based processing to make something like this happen, but most likely won't. The best we could hope for is pre-programmed voice commands with canned responses, though a lot of cool stuff could still be done with that given enough developer effort.
