[REL] sh4speech - voice command for SH4

[ferg]

Originally Posted by billko

Sounds like it could be related to the special CSV files for your specific mod. Since SH4Speech hasn't been udated in a while, they may have become outdated. Try asking around to see if anyone has made some revised ones.

You're basically right, Billko. The problem does not affect "stock", but it does affect "trigger maru" (thus my own modified "trigger maru" also has the problem). However, I stand by my description of it as a bug since it occurs with an included data set. As far as I know, I haven't modified anything in the "trigger maru" set. I'm going to check that right now by snagging the file again and overwriting the "trigger maru" files to confirm it.

Edit: Confirmed. The problem exists in the provided "trigger maru" files. I'll see if I can figure it out. Thanks for the clue that it works with the stock data! With any luck I should be able to fix it.

Edit: Turns out the problem affects the script anytime the resolution is set to 1440x900...

[billko]

Originally Posted by ferg

You're basically right, Billko. The problem does not affect "stock", but it does affect "trigger maru" (thus my own modified "trigger maru" also has the problem). However, I stand by my description of it as a bug since it occurs with an included data set. As far as I know, I haven't modified anything in the "trigger maru" set. I'm going to check that right now by snagging the file again and overwriting the "trigger maru" files to confirm it.

Edit: Confirmed. The problem exists in the provided "trigger maru" files. I'll see if I can figure it out. Thanks for the clue that it works with the stock data! With any luck I should be able to fix it.

Mucking around inside the changelog shows that the last version of TMO supported was 1.6. Also, the last version of SH4Speech seemed to have a different way of handling the TMO CSV files and warned of compatibility issues. (see first page, this topic)

Again, people here probably have updated CSV files that could save a lot of headache

I hope this helps.

Bill

[ferg]

Originally Posted by billko

Also, the last version of SH4Speech seemed to have a different way of handling the TMO CSV files and warned of compatibility issues. (see first page, this topic)

First page of this topic does not contain "TMO" or the phrase "trigger maru". I can't seem to find the reference. I'm not sure what you mean about a different way of handling the TMO CSV files.

There are slight differences between running sh4Speech in "stock" and "trigger maru" modes in the code, but nothing that looked like it would break anything. Looking at the source, I only noticed that torpedo tube handling is somehow different (probably to handle different tube cycling in TM). I've got tube cycling working well enough for me at this point, so that is not a concern.

Again, people here probably have updated CSV files that could save a lot of headache

Totally, but I'm not sure anyone has put up with the headache and done it. If they have, why is there nothing on minsc's download page or in this thread? A Google search for "sh4speech tmo 1.9" finds only my own post asking if anyone has done an update. I tried the more general "sh4speech tmo" too, but no dice. If I can get it working to my satisfaction, I'll happily make my updates available to others.

[billko]

Originally Posted by ferg

First page of this topic does not contain "TMO" or the phrase "trigger maru". I can't seem to find the reference. I'm not sure what you mean about a different way of handling the TMO CSV files.

There are slight differences between running sh4Speech in "stock" and "trigger maru" modes in the code, but nothing that looked like it would break anything. Looking at the source, I only noticed that torpedo tube handling is somehow different (probably to handle different tube cycling in TM). I've got tube cycling working well enough for me at this point, so that is not a concern.



Totally, but I'm not sure anyone has put up with the headache and done it. If they have, why is there nothing on minsc's download page or in this thread? A Google search for "sh4speech tmo 1.9" finds only my own post asking if anyone has done an update. I tried the more general "sh4speech tmo" too, but no dice. If I can get it working to my satisfaction, I'll happily make my updates available to others.

Darn it. If I had that mod I would do the same.

[Chromius]

Has anyone got some modded files that work with RFB 2.0 (1680x1050) just trying to save myself time.

[ferg]

OK, well, this "damage control team" problem is a strange one. It turns out it has nothing to do with the CSV files. I thought the problem wasn't appearing in stock until I remembered that I had left the stock batch file at 1024x768. When I changed it to 1440x900, it evidenced the exact same behaviour. There is a bug in the script itself for that resolution.

I don't know Perl and I don't have a runtime installed on my computer, so it's a bit hard for me to debug. Nevertheless, I have a powerful suspicion that the script is going awry in the MouseMoveAbsPix subroutine. It enters an until() loop on line 811, and I think the condition never evaluates true.

But why it's happening, why I've only seen it triggered with this one command, I have no idea. It's possible that it occurs with other commands as well. I probably haven't used every single command defined. Another command with the same co-ordinates ("Aim For Weapons", ID=543) works fine.

[billko]

Originally Posted by ferg

OK, well, this "damage control team" problem is a strange one. It turns out it has nothing to do with the CSV files. I thought the problem wasn't appearing in stock until I remembered that I had left the stock batch file at 1024x768. When I changed it to 1440x900, it evidenced the exact same behaviour. There is a bug in the script itself for that resolution.

I don't know Perl and I don't have a runtime installed on my computer, so it's a bit hard for me to debug. Nevertheless, I have a powerful suspicion that the script is going awry in the MouseMoveAbsPix subroutine. It enters an until() loop on line 811, and I think the condition never evaluates true.

But why it's happening, why I've only seen it triggered with this one command, I have no idea. It's possible that it occurs with other commands as well. I probably haven't used every single command defined. Another command with the same co-ordinates ("Aim For Weapons", ID=543) works fine.

wow, i'm not sure why it would be resolution-related because it's all just calculating ratios for mouse movement, but there's no denying there's something amiss! i'll have to think about that one. hopefully minsc will be by soon to bring to light the solution

[minsc_tdp]

Ok guys this seemed like the best place to post this.

If anyone is having trouble with the new voice command system for SH5, and used my shSpeech, and you want to try it with SH5, check out this webpage as it has updated files for SH5.

http://knepfler.com/shSpeech

These were created by 7Infanterie19 not me, so please no emails! Unfortunately I have no time to support this or do really anything further on it, but 2.54 is quite stable. I haven't tested these new CSVs, so, good luck everyone!

Now, what's all this about damage control... I promise to try to look when I get a chance!

[minsc_tdp]

Originally Posted by ferg

Problem solved. TMO does indeed transpose the dials. Using a screenshot from the stock game, I figured the TMO dial centres to be 35 pixels higher. By adjusting the values in dials.csv, I got it to work just fine. I think I'm going to enjoy sh4Speech very much once I work out all the TMO incompatibilities.

Thanks for your help Billko.

If anyone is so inclined as to organize changes like this into a package, and call it something like MODIFIED CSVS FOR RFB 2.0.ZIP and PM me, I'll send you my email address and distribute them on the main page for all. I'm sorry I haven't been able to keep up but glad you guys are managing to make it work by reading the docs and digging into the CSVs. I did my best to make it community-sustainable.

[minsc_tdp]

Originally Posted by ferg

If the ID is between 19 and 28 inclusive, a mouse move is added. Because minsc kindly included the source, I think I can fix this for myself if I download the necessary compiler; but I think a fix to the official version would benefit all. It's also possible that I'm completely missing something. Maybe there's a reason for this mouse move? I can't imagine what it is.

This is definitely a major regret of mine. It saddens me to think I have code that actually looks at specific IDs. I think according to the comment above:

# Move to center area to prevent obscuring the screen for certain commands

The idea was to move the mouse cursor to the center as leaving it at the bottom caused a tooltip or something which would then obscure part of the screen.

This may be solvable in general in the CSV. If you see ANY reference to a specific ID number in the voice.pl script, where it looks like it's handling it special and it's screwing things up, the solution may be simple: change the ID to something else in the CSV. Most of the ID numbers are arbitrary and ignored by the code, or used only for linking commands and stuff like that. Try a really crazy ID number like 94293.

Yes you could download the compiler (all you'd need is ActiveState Perl and perl2exe.exe) and tweak the Perl if you like but finding a solution in the CSVs is always preferred since everyone else can implement them more easily I would think.

[minsc_tdp]

Originally Posted by ferg

Totally, but I'm not sure anyone has put up with the headache and done it. If they have, why is there nothing on minsc's download page or in this thread? A Google search for "sh4speech tmo 1.9" finds only my own post asking if anyone has done an update. I tried the more general "sh4speech tmo" too, but no dice. If I can get it working to my satisfaction, I'll happily make my updates available to others.

Nobody has ever bundled up an updated set of CSVs for versions of Trigger Maru past 1.6, hence there's nothing on the webpage about it. You might be the first to have bothered. PM me if you manage to get it working and have all your CSVs organized into a zip and I'm happy to post to the site.

That damage control bug is really scary. Seems like it is locking up - I'd bet voice.exe is running 100% CPU when this happens. Very ugly.

[minsc_tdp]

Originally Posted by ferg

OK, well, this "damage control team" problem is a strange one. It turns out it has nothing to do with the CSV files. I thought the problem wasn't appearing in stock until I remembered that I had left the stock batch file at 1024x768. When I changed it to 1440x900, it evidenced the exact same behaviour. There is a bug in the script itself for that resolution.

I don't know Perl and I don't have a runtime installed on my computer, so it's a bit hard for me to debug. Nevertheless, I have a powerful suspicion that the script is going awry in the MouseMoveAbsPix subroutine. It enters an until() loop on line 811, and I think the condition never evaluates true.

But why it's happening, why I've only seen it triggered with this one command, I have no idea. It's possible that it occurs with other commands as well. I probably haven't used every single command defined. Another command with the same co-ordinates ("Aim For Weapons", ID=543) works fine.

It's very strange that it happens only for that one command. I don't recall there being anything special about damage control.

Taking a quick look, I see it has an ID of 55 in the Stock CSV set. I also see some trigger maru specific code that does a general text inspection at the CSV set name to determine if it has "trigger maru" in it, and if it does, it treats 55 special during the tube specific code, which is weird because in my set, trigger maru damage management is still 55, in other words, 55 wasn't repurposed to tubes or anything in tm. So there could be a problem there.

One fix might be to make sure your CSV set name as referenced in the launch bat file doesn't say "trigger maru" which would cause it to be treated as stock (of course if you're used to running in TM mode, that might cause a bit of a shock.) Another option might be to change all instances of ID 55 to some other obscure number as mentioned above. Whenever you suspect something weird about the ID, change it, and be sure to change it in every CSV it might be referenced in, in both the ID column and the PARENT column.

Generally speaking most testing was done in stock, and TM was kind of tacked on and was never really fully vetted as well as it could have been.

Hope that's clear enough and leads you to a solution... sorry I can't even test this as I don't have a working microphone at the moment. I do on my laptop, if I have time I will try tomorrow!

[ferg]

Sorry, I am just logging off and heading to bed or I'd write a proper reply, Erik. I'll get to that tomorrow or Thursday depending. Thanks for taking the time to post.

I just wanted to share that I installed ActivePerl and enabled the debug output you had in that until loop. This is what I found interesting:

Mouse clicking at 336,871 for Damage Control Team (id=572)
17476==15292&&59637==63423
17039==15292&&60394==63423
16602==15292&&61151==63423
16165==15292&&61909==63423
15728==15292&&62666==63423
15291==15292&&63423==63423
14855==15292&&64181==63423
...and off it goes into an endless loop...

The second to last line is where the until loop is intended to exit, but of course it doesn't because 15291 is not 15292. I only looked at it for a minute or so, but off the cuff it looks like a rounding error to me.

[billko]

Originally Posted by ferg

Sorry, I am just logging off and heading to bed or I'd write a proper reply, Erik. I'll get to that tomorrow or Thursday depending. Thanks for taking the time to post.

I just wanted to share that I installed ActivePerl and enabled the debug output you had in that until loop. This is what I found interesting:

Mouse clicking at 336,871 for Damage Control Team (id=572)
17476==15292&&59637==63423
17039==15292&&60394==63423
16602==15292&&61151==63423
16165==15292&&61909==63423
15728==15292&&62666==63423
15291==15292&&63423==63423
14855==15292&&64181==63423
...and off it goes into an endless loop...

The second to last line is where the until loop is intended to exit, but of course it doesn't because 15291 is not 15292. I only looked at it for a minute or so, but off the cuff it looks like a rounding error to me.

hey nice catch!

[kraszus]

Hi. First, thanks for a great mod

I play Trigger Maru variant using 1680x1050 resolution. I have edited the main.cfg and the shspeech.bat accordongly:-

@cd dist
voice 1680x1050 "trigger maru" 1 right sh4

Everything works great except Heading Change requests and Depth Change requests. Eg. If I say 'heading left 10 degrees' it sets a heading about 60 degrees left instead of 10 degrees. Similarly with depth if asked to set a depth to 40meteres, it sets a depth to 80 or 90 metres.

I can see the mouse click on the dial and it is clicking in the wrong place.

Have I edited the files correctly?

PS. I should probably mention that I have also downloaded the SH3 variant and it works fine with the dials.