iPhone project advice?

Started by
1 comment, last by swiftcoder 15 years, 9 months ago
Hello, As background, I work for Annosoft. We developed Speech software for lipsync. Dreaming about porting our realtime lipsync to iPhone and other mobile platforms. Whether there are is market, who knows. It sounds pretty cool, talking characters. Is it doable... The speech system is floating point math intensive. n-cube algorithm, with single precision, and one spot requiring double precision math to avoid numeric underflow. floating point, as I understand it, was always a mobile-no-no. Is that still true? If floating point is decent, performance should be decent and visa-versa... The code is very portable. I'm guessing compilation won't be horrible. The next question. For a demonstration product, It would be cool to show one of our avatars and have him lipsync to the user as he/she talks. Like our win32 demo. Does anyone know if you have access to audio from the mic, with low latency."Maybe press a button to run" - lipsync while a button/key is down. Anyway, that's a mouth full, and makes some assumptions... any advice is appreciated? Thanks and Regards, Mark Zartler http://www.annosoft.com Annosoft Really Great Lipsync
Mark Zartler
Advertisement
From the iPhone documentation:

"iPhones and iPod touches support two instruction sets, ARM and Thumb. Xcode uses Thumb instructions by default because using Thumb typically reduces code size by about 35 percent relative to ARM. Applications that have extensive floating point code might perform better if they use ARM instructions rather than Thumb. You can turn off Thumb for your application by turning off the
Compile for Thumb build setting."

Which seems to indicate that you might be okay if floating point calculations are your only concern. As for the mic, I don't see why you wouldn't use the same API features for recording audio.
Check out my current project on Kickstarter: Genegrafter
Quote:Original post by annosoft
The speech system is floating point math intensive. n-cube algorithm, with single precision, and one spot requiring double precision math to avoid numeric underflow. floating point, as I understand it, was always a mobile-no-no. Is that still true?
Several developers have noted that the iPhone's floating point performance is surprisingly good for a mobile device. How well that stacks up against a PC, I have no idea.

Tristam MacDonald. Ex-BigTech Software Engineer. Future farmer. [https://trist.am]

This topic is closed to new replies.

Advertisement