Rockbox

This is the bug/patch tracker for Rockbox. Click here for more information.

Quick links: Bugs · Patches · Rockbox frontpage

Tasklist

FS#11887 - Fix English TTS Pronounciation

Attached to Project: Rockbox
Opened by Christian Mertes (mudd1) - Tuesday, 18 January 2011, 00:11 GMT+2
Last edited by Dominik Riebeling (bluebrother) - Sunday, 23 September 2012, 21:39 GMT+2
Task Type Patches
Category User Interface
Status Closed
Assigned To No-one
Player Type All players
Severity Low
Priority Normal
Reported Version Rbutil SVN
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Private No

Details

There are a lot of menu entries that aren't pronounced as they are supposed to. The following patch will hurt your eyes but it will stop the harm done to your ears. I hope I got all bad TTS occurrences, just let me know if I didn't. I tried with festival using the cmu_us_slt_arctic_clunits voice. I would be glad if anybody could try using other engines.
   fix_english_tts_pronounciation.patch (2.9 KiB) 
 apps/lang/english.lang |   34 +++++++++++++++++-----------------
 1 file changed, 17 insertions(+), 17 deletions(-)

This task depends upon

Closed by  Dominik Riebeling (bluebrother)
Sunday, 23 September 2012, 21:39 GMT+2
Reason for closing:  Out of Date
Additional comments about closing:  Please put word adjustments in tools/voice-corrections.txt since they are TTS specific.
Comment by Nils Wallménius (nls) - Tuesday, 18 January 2011, 09:11 GMT+2
Other corrections like this are done in voice.pl since different ones are needed for different tts's, so you should move this in there instead.
Comment by Christian Mertes (mudd1) - Tuesday, 18 January 2011, 10:29 GMT+2
Oh ok, didn't know that. I will try to put it in there. Do you have any idea why I had to do the Plugins->Plug-ins transformation manually? Shouldn't the 'i' after the s// operator in that Perl code have made it case-insensitive?
Comment by Nils Wallménius (nls) - Tuesday, 18 January 2011, 14:30 GMT+2
Ah, you use rbutil for voice generation, rbutil doesn't use the voice.pl script so that's why. I don't know rbutil so have no idea if it could use the script.
Comment by Dominik Riebeling (bluebrother) - Saturday, 29 January 2011, 19:48 GMT+2
See also  FS#11913  which is an attempt to make the corrections to strings for the TTS engines available to other tools.
Comment by Christian Mertes (mudd1) - Monday, 31 January 2011, 15:35 GMT+2
I had a look at the makefile and it seems to me that "make voice" does use the voice.pl script. I welcome the effort done in  FS#11913  but to my understanding the rules should already work for me. Any idea why they don't?
Comment by Nils Wallménius (nls) - Monday, 31 January 2011, 16:02 GMT+2
make voice does use voice.pl rbutil does not, bluebrother's patch is a first step to mkaing rbutil able to use these corrections.
I though you used rbutil for voice generation since you have set "Rbutil SVN" as the reported in version, is that not the case?
Comment by Dominik Riebeling (bluebrother) - Monday, 31 January 2011, 17:16 GMT+2
Judging from your patch you're modifying much more strings (like Encoder -> En-coder) than the corrections voice.pl currently does. My guess would be that the corrections in fact _did_ work for you, it's just that they aren't sufficient for the TTS engine you used.

Just as an example: I'm using a voice file created with the voice "Alex" on OS X, and even without any corrections (as created with Rockbox Utility) the voice file sounds good. A voice file created with the default voice of Windows XP is hardly bearable. This clearly shows that the corrections are heavily TTS specific.
Comment by Christian Mertes (mudd1) - Monday, 31 January 2011, 17:30 GMT+2
Sorry, I'm using an SVN built but without the Rockbox Utility. I think "Rbutil SVN" was just the closest match to "SVN" I saw.

I did change much more than voice.pl does but for example the transformation "plugin→plug-in" was something I came up with before I learned that voice.pl does the exact same thing for festival (which I use). It nevertheless improved the generated voice. Any idea why that might have been the case?

Loading...