RE: Speech output to file

If this is a one time job instead of a packaged app, you may also want to consider /usr/bin/say (see the man page), which offers a lot of output options conveniently packaged.

I will give “say” a shot. I thought that rendered in real-time, which is one of my concerns. I need to be able to quickly as possible produce files for listening.

>>The recommended API nowadays is to use the kSpeechOutputToFileURLProperty…

I believe I understand. I will go and make it happen.

Thank you,

Waverly

From: speech-dev-bounces+waverly.edwards=email@hidden [mailto:speech-dev-bounces+waverly.edwards=email@hidden] On Behalf Of Matthias Neeracher
Sent: Wednesday, May 18, 2011 4:22 PM
To: email@hidden
Subject: Re: Speech output to file

On May 18, 2011, at 21:45 , Edwards, Waverly wrote:

I am looking to output just short of 1000 of pages of text to file. I have not decided if the outoput will be one file per page or one file per chapter.

Are there any example of how to output speech to a file.

The recommended API nowadays is to use the kSpeechOutputToFileURLProperty or kSpeechOutputToExtAudioFileProperty selectors to SetSpeechProperty(). Traditionally there are the soOutputToFileWithCFURL & soOutputToExtAudioFile selectors to SetSpeechInfo, as Jay Kreibich mentioned. The ...ToExtAudioFile variant gives a lot more control, but is more intricate to set up.

If this is a one time job instead of a packaged app, you may also want to consider /usr/bin/say (see the man page), which offers a lot of output options conveniently packaged.

Is there a way to render way to render speech output faster than realtime?

All of these methods will render faster than realtime. If you have a lot of cores and multiple output files, you can use separate processes (separate threads/channels alone will not suffice). MacinTalk uses up to 2 cores per process, so on an 8 core machine, you might be able to obtain another 4x speedup using separate processes.

I would imagine that many pages of text taking an enormous amount of time to render.
If there is a way to render the output to a file, is it possible the speech manager reference a pronunciation dictionary for words the speech manager does not know?

The API to do so exists, but (at least in MacinTalk), it's not hooked up. You could always preprocess the string, replacing specific words with [[inpt PHON]]...[[inpt TEXT]]

--
Matthias Neeracher
Senior Software Engineer, Accessibility Team, Apple Inc

References:
	>Speech output to file (From: "Edwards, Waverly" <email@hidden>)
	>Re: Speech output to file (From: Matthias Neeracher <email@hidden>)