Hard Light Productions Forums
Modding, Mission Design, and Coding => Help Wanted/Offering Help => Topic started by: Iain Baker on July 02, 2018, 10:51:03 am
-
Hi there!
I’m sure we have all cringed at how poorly the text-to-speech system works in game (and in general – YouTube videos, I’m looking at you).
However, there are ways of improving it, and done correctly it can actually be quite effective. Feel free to take a look at my website to see examples. https://iainbakerfreelance.co.uk/
If anyone would like me to help with the text to speech optimisation of their mod please get in touch.
One question, is the text-to-speech software narrating the text that is on screen, or is it referencing a file somewhere? If so, would altering this file alter the text that appears on screen? I ask as optimising text for text-to-speech narration involves manually altering the text into a format that the software is better able to understand. This results in narration that sounds pretty decent, but if someone were to read the text it would look like gobbledygook ;-)
Ideally you would want the on screen text to be referencing one unaltered file, and the narration referencing a separate optimised file - or a pre-recorded MP3 file. You would then have on screen text that is readable and narration that is listenable.
Hope to hear from you soon :)
-
Unfortunately it's reading straight off the screen.
-
You can have the in-mission dialog use different text than what you're using for text-to-speech, but that would entail pre-recording the audio and bundling it with your mod.
-
And the benefit from having audio different from text would be... higher quality?
-
Yes, it sounds like you can give the TTS something closer to a phonetic script to read so it doesn't get confused.
-
Problem is, sometimes FREDders use variables or player's name in a message content, which can't be prerecorded.
-
Can the TTS be altered to use the file instead? And then to have a preference for a TTS optimized file if it exists?
Or perhaps modify FSO to pass the contents of an optimized file instead of the normal screen content if it exists?
Not quite sure how Windows TTS works or how much control there is over its behavior. If it's Windows just reading off the screen, then how come it doesn't read off the ship types you're targeting? Or are those displayed as images?
-
Not quite sure how Windows TTS works or how much control there is over its behavior.
Quite a lot of control over what's said, actually.
If it's Windows just reading off the screen
It isn't. We specifically feed message text to the SAPI.
-
NB - I can record the TTS as an MP3 or WAV file, which you can then use in your mods in lieu of 'proper' people narration :)
Samples can be heard by following the link: https://iainbakerfreelance.co.uk/text-to-speech-optimisation/