Guest Posted June 8, 2021 Posted June 8, 2021 On 6/4/2021 at 1:28 PM, Roggvir said: Just in case anybody would want to use that script. There was an issue that it creates some files with having the TXT extension in uppercase, which xVASynth (currently 1.37) refuses to load. So, here is the script with that problem rectified. split.py 4.5 kB · 1 download Maybe I am doing something wrong, but I've noticed that dialogue lines where the text contains comas, get's broken out into more columns in the final output. So, I've taken to replacing comas in the initial Tab delimited CSV. I've found that XVA can handle a period like a coma. In fact, XVA handles comas like shit anyway. I don't know much.....Fuck, I don't know anything about python. But I tweaked with one of the scrubber lines in your script where if in the NPC dialogue, it finds a coma, it replaces it with a period. text = f.read().replace(',','.').replace('\n','').replace('\r','') # replace is a tiny bit faster than the regex, even if we call it twice It looks like it worked.....Again, not sure what I have done here, but I'll try to take it as far as the xva batch file.
Herowynne Posted June 10, 2021 Posted June 10, 2021 On 5/30/2021 at 5:18 PM, finalfrog said: I generated voice packs for all the voices which xVASynth supported at the time I released them. Since then the author has released a number of new voice models, but sadly MaleEflHaughty isn't one of them. MaleEflHaughty just released for xVASynth ! Yay! SKVA Synth - xVASynth at Skyrim Special Edition Nexus - Mods and Community (nexusmods.com)
Roggvir Posted June 14, 2021 Posted June 14, 2021 On 6/8/2021 at 2:36 PM, bishlapped said: Maybe I am doing something wrong, but I've noticed that dialogue lines where the text contains comas, get's broken out into more columns in the final output. So, I've taken to replacing comas in the initial Tab delimited CSV. I've found that XVA can handle a period like a coma. In fact, XVA handles comas like shit anyway. I don't know much.....Fuck, I don't know anything about python. But I tweaked with one of the scrubber lines in your script where if in the NPC dialogue, it finds a coma, it replaces it with a period. text = f.read().replace(',','.').replace('\n','').replace('\r','') # replace is a tiny bit faster than the regex, even if we call it twice It looks like it worked.....Again, not sure what I have done here, but I'll try to take it as far as the xva batch file. That is weird, it shouldn't need any such changes. The input file is supposed to be a TAB delimited file - as exported by Creation Kit - so there shouldn't be any problem with commas. I see the script i posted is significantly different from the one i am actually using, not sure how the hell did that happen. So, here is the actual current version i am using, which at least on my side has no problems with any commas and such. split.py
Guest Posted June 14, 2021 Posted June 14, 2021 Yeah, The XVA Utility should have been formatted to use TAB Delimited when working with converted dialogue output text files. But thanks for this.
Guest Posted June 15, 2021 Posted June 15, 2021 On 6/14/2021 at 12:10 AM, Roggvir said: That is weird, it shouldn't need any such changes. The input file is supposed to be a TAB delimited file - as exported by Creation Kit - so there shouldn't be any problem with commas. I see the script i posted is significantly different from the one i am actually using, not sure how the hell did that happen. So, here is the actual current version i am using, which at least on my side has no problems with any commas and such. split.py What would be involved with altering your script to work with FO4 exports? Is it worth taking a look? The CSV column format is different. I can try to figure it out. I know a tiny bit about scripting in general, but nothing about PY specifically. If I interpret your script, you are checking for a number of rows, and then reading the contents from a few of them.
Roggvir Posted June 15, 2021 Posted June 15, 2021 3 hours ago, bishlapped said: What would be involved with altering your script to work with FO4 exports? Is it worth taking a look? The CSV column format is different. I can try to figure it out. I know a tiny bit about scripting in general, but nothing about PY specifically. If I interpret your script, you are checking for a number of rows, and then reading the contents from a few of them. It should be easy to adjust it for Fallout 4. You need to change in which column the script looks for the voice type, which is on line 74: voice = row[5] ...change 5 to whichever column in the FO4 export contains the voice type (keep in mind this is a zero-based index, so 5 is actually 6th column). You need to change the columnsCount != 24 condition on line 62, to reflect the number of columns the FO4 export has (i don't know how many that is, you will have to count them yourself). Same on line 75 where the condition len(row) != 24 checks number of items in the "row" list. Also on line 75, note the condition row[0]!='FFFFFFFF', which is a simple check to help ensure the CSV is formatted properly, relying on the Skyrim CK dialogue export having FFFFFFFF in first column on every line, but i do not know if Fallout 4 export is the same, maybe you need to remove this condition (it is not strictly necessary anyway, as the other checks for a proper format should be enough). On line 101 you probably need to adjust the column index used to get the response text, as that may be in a different column in the FO4 dialogue export, again, keep in mind the index is zero-based: line = normalizeText(row[20]) That should do it, that be all there is that may neeed some changes.
Guest Posted June 18, 2021 Posted June 18, 2021 On 6/1/2021 at 2:42 PM, Roggvir said: I have a problem with doing Submissive Lola dialogue exports from Creation Kit. The resulting export is missing a lot of lines. The cause of it seems to be a topic info condition like... ...because the alias is only filled at real-time by a script, the condition returns false while in Creation Kit, and because CK apparently evaluates this condition when building list of voice types for this line, it ends up as having no available voice. I am trying to find a reasonable and systemic solution to this. I could set the alias to some NPC, which will result in the line being included in the export - BUT only for the voice that particular NPC is using, so that isn't a systemic solution at all. I could replace the condition with another one, checking the subject's voice type against formlist of all voice types i know can be used by any NPC that can become the "Owner" - BUT there are MANY topic infos where this condition can be, so this is just plain insane. So far, the best i was able to come up with is: Create a dummy "Owner" NPC, create and give it a new "OwnerVoice" voice type, and set this NPC into that "Owner" alias (that will ensure the line appears in the export using my special "OwnerVoice"), and do the export. Then create duplicates of any CSV lines using my special "OwnerVoice" for all the voice types i know the "Owner" to be. But isn't there a better solution? There must be a better way, right? This is just one case out of many where something like this can result in lines not getting exported, surely there must be some intelligent way of handling this problem? With this particular mod, I have tried what you suggested. I took Jenassa, and duplicated her and re labeled her as "Owner" then I took Fem DE voice and duplicated it as OwnerVoice type.....It all seems to have worked and I was able to assign the Owner Alias to an actual NPC. This allowed the Dialogue export to see all the dialogue. But it's odd, that the only voice type that it allowed for the dynamic dialogue (such as when the owner says "You actually did it. You always do as your are told" ......It came back as one viable voice type. FemaleEvenToned. No idea why. But at least all the text and files came back in the export. From there I tweaked the batch. It's still processing my 1st test run of voice gen for this mod. So far, a random sampling shows mediocre results.......Quality wise. I may go back and play with the text on certain words to get better results. But rather than use the skyrim voices, I am generating the audio using the CP_V_Female voice. SubLolaSample.wav
Roggvir Posted June 18, 2021 Posted June 18, 2021 1 hour ago, bishlapped said: With this particular mod, I have tried what you suggested. I took Jenassa, and duplicated her and re labeled her as "Owner" then I took Fem DE voice and duplicated it as OwnerVoice type.....It all seems to have worked and I was able to assign the Owner Alias to an actual NPC. This allowed the Dialogue export to see all the dialogue. But it's odd, that the only voice type that it allowed for the dynamic dialogue (such as when the owner says "You actually did it. You always do as your are told" ......It came back as one viable voice type. FemaleEvenToned. No idea why. But at least all the text and files came back in the export. From there I tweaked the batch. It's still processing my 1st test run of voice gen for this mod. So far, a random sampling shows mediocre results.......Quality wise. I may go back and play with the text on certain words to get better results. But rather than use the skyrim voices, I am generating the audio using the CP_V_Female voice. SubLolaSample.wav 229.08 kB · 0 downloads I later described the way how i prepare the data/plugin for the CK Dialogue Export here: https://www.loverslab.com/topic/168093-xvasynth-discussion/?do=findComment&comment=3401236 Its more complicated than what you did, but it will result in what seems to be the correct range of voice types.
Guest Posted June 19, 2021 Posted June 19, 2021 1 hour ago, Roggvir said: I later described the way how i prepare the data/plugin for the CK Dialogue Export here: https://www.loverslab.com/topic/168093-xvasynth-discussion/?do=findComment&comment=3401236 Its more complicated than what you did, but it will result in what seems to be the correct range of voice types. I def did something wrong.....I seem to have all the files and lips, generated but I can't seem to get the mod to recognize them. Plus I have the added complication of trying to add a custom voice type. Not sure that will work. At least not in how I did it. I'll read up on this, thanks.
Roggvir Posted June 19, 2021 Posted June 19, 2021 4 hours ago, bishlapped said: I def did something wrong.....I seem to have all the files and lips, generated but I can't seem to get the mod to recognize them. Plus I have the added complication of trying to add a custom voice type. Not sure that will work. At least not in how I did it. I'll read up on this, thanks. When you wrote "can't seem to get the mod to recognize them", you mean when in game, the game is not using the voice/lip files? To generate the LIP files, you had to rename the "plugin" folder from "submissivelola_est.esp" to "submissivelola_est.esm", right? ...at least without that, the LIP generation doesn't work for me. So, maybe the game cannot find the files, because you forgot to rename it back to "submissivelola_est.esp"? Regarding custom voice type... Sure, you can do that, but do you have some NPC set to actually use the new custom voice type? Or, if for whatever reason you prefer the CP V voice, you can use that without creating a custo mvoice type if you just put the files into a subfolder meant for some existing voice type used by whichever follower you intend to use (though i would find it annoying when the follower talks using its normal voice when saying some vanilla game lines, and then suddelnly switches to a different voice when saying stuff from the mod). Btw. a little tip you can try to get better/cleaner sounding voice: You can try prefixing every text line with ", " (comma followed by space), so "hello, how are you?" becomes ", hello, how are you?" This CAN make many lines sound much better, though it may depend on the voice type - sometimes even if you get better/cleaner sounding voice, it may butcher pronunciation of some words even though without the comma prefix the same words would get pronounced correctly. When i was doing the FemaleDarkElf voice type, i generated all lines twice - once normally, and once with the comma prefix, and then for every single line i picked the result which sounded better, and if none was good enough, i tweaked the line manually. Without the comma prefix, almost all lines sounded awful, and when using the comma prefix, about 50% of the lines were good enough out of the box.
Guest Posted June 19, 2021 Posted June 19, 2021 5 hours ago, Roggvir said: When you wrote "can't seem to get the mod to recognize them", you mean when in game, the game is not using the voice/lip files? To generate the LIP files, you had to rename the "plugin" folder from "submissivelola_est.esp" to "submissivelola_est.esm", right? ...at least without that, the LIP generation doesn't work for me. So, maybe the game cannot find the files, because you forgot to rename it back to "submissivelola_est.esp"? Regarding custom voice type... Sure, you can do that, but do you have some NPC set to actually use the new custom voice type? Or, if for whatever reason you prefer the CP V voice, you can use that without creating a custo mvoice type if you just put the files into a subfolder meant for some existing voice type used by whichever follower you intend to use (though i would find it annoying when the follower talks using its normal voice when saying some vanilla game lines, and then suddelnly switches to a different voice when saying stuff from the mod). Btw. a little tip you can try to get better/cleaner sounding voice: You can try prefixing every text line with ", " (comma followed by space), so "hello, how are you?" becomes ", hello, how are you?" This CAN make many lines sound much better, though it may depend on the voice type - sometimes even if you get better/cleaner sounding voice, it may butcher pronunciation of some words even though without the comma prefix the same words would get pronounced correctly. When i was doing the FemaleDarkElf voice type, i generated all lines twice - once normally, and once with the comma prefix, and then for every single line i picked the result which sounded better, and if none was good enough, i tweaked the line manually. Without the comma prefix, almost all lines sounded awful, and when using the comma prefix, about 50% of the lines were good enough out of the box. I think I broke the mod when I started messing with adding a new voice type. It's not that the voice files aren't seen. I did revert the files and folders all back to ESP. (Yeah, that one got me before) I am guessing I created a disconnect between my follower and either the alias or the voice type. It's like the whole mod won't start. I was going to start over based on your previous recommendation. 1st use a Skyrim based voice type and figure out how it's supposed to work before I get creative.
Roggvir Posted June 19, 2021 Posted June 19, 2021 10 hours ago, bishlapped said: It's like the whole mod won't start Maybe it has nothing to do with your custom voice type (or not only with that), maybe its because of how your custom follower is configured - i suggest you ask in the Lola support topic, whether there are any specifics to consider when adding a new follower.
Recommended Posts
Archived
This topic is now archived and is closed to further replies.