MD's Bethesda Voice Tools

[MellowDrama]

View File

MD's Bethesda Voice Tools

---

The .pas script in downloads is just a backup copy. The installer will make a new with the directory already in it. Use that one obviously.

 

I'll keep it short.

I keep forgetting how Skyrim's voice conversion stuff works. with the fuzzes and the lips and the wavs and the mp3s and the naming convention so I made this.

the first bat runs an installer. After that you should be able to run the SeEdit script .pas on a specific dialogue topic or on the entire esp and it will automatically open MD's Skyrim Voice Tools which opens Qwen3 which uses the included 30 seconds of soundbites from the actual game to generate all your lines. Then it converts to .fuz with actual .lip and renames then to the correct format and places them in the correctly named folder. 
All you need to do is put it in the Data folder afterwards.

 

There's an option to generate multiple lines at once with slightly different settings. 0.75-0.8 seems to work the best for me but I will caution you if you are CUDA Cucked like I am CPU generation does work but it will take half a minute per line. There's a 126 voices included in skyrim and 705 in FO4 and another ~700 in FOLONwhich will take weeks if you have a very large mod and you do multiple tries for each line. You can turn off voices if you're sure they won't use your lines.

You can safely close it, it will continue where you left off if you run the .pas again in SeEdit.

I made this and it works for me, it should work for you but if it doesn't let me know. Also let me know if there's anything confusing about the process. 

 

Supported languages:

Skyrim

-English
-French
-German
-Italian
-Japanese
-Russian
-Spanish

Fallout 4:

-English
-French
-German
-Italian
-Japanese

-Spanish

Fallout London

-Bri'ish

 

You can add custom voices to for example add lines to modded followers etc. All you need is a 30 second mp3 file of stitched together lines and an optional transcript of what is said.

(It will only add that voice to the currently selected game and language)

 

These are the only languages that have actual audio in Skyrim and fallout 4 AND are supported by Qwen3. Sorry to all the Norwegians, you'll just have to learn a real language.

 

If you get a warning about Sox, Just go on ahead and ignore that.

These are some examples

 MaleOrc1

 

 MaleYoungEager

 

Cicero

   

 

Japanese

  

 

German 

  

 

Russian

 

---

• Submitter
  MellowDrama
• Submitted
  07/27/26
• Category
  Regular Mods
• Requirements
• Regular Edition Compatible
  Yes
• Install Instructions

  unzip and put it in. 😛

 

[Pavlovian Afterdark]

The Cicero sample line made me chuckle. Thank you for sharing.

[bigbasi]

Are you kidding me... I just spend the day starting to program my own tool, trying to create better results than with xvasynth 😉

 

congrats, for beating me to it 😉

 

Installer ran fine without errors, Path is configured

 

Anyway, I just ran into an error:

 

MD's Skyrim Voice Tools path is not configured in ExportDialogueVoiceJobsAndLaunchTools.pas.
Run the full installer, or edit VoiceGeneratorPath near the top of this script.

 

const
  {
    ...

  }
  VoiceGeneratorPath = 'E:\Nefaram 17.0.5\Tools\MD''s Skyrim Voice Tools\skyrim-audio-tools';

 

function ResolvedVoiceGeneratorPath: string;
begin
  Result := WithoutTrailingSlash(StripQuotes(VoiceGeneratorPath));

  if Result = 'E:\Nefaram 17.0.5\Tools\MD''s Skyrim Voice Tools\skyrim-audio-tools' then
    Result := '';
end;

Edited July 27 by bigbasi

[necol]

Hi! First, thank you for the great tool.
I'm testing Qwen3-TTS-0.6B-CustomVoice with Skyrim Voice Tools.
English works perfectly.
However, when I set Language = Russian and use Cyrillic text, the generated speech is unintelligible (gibberish).
Is Russian currently supported by the backend/model? If yes, what language value should be used (Russian, ru, ru-RU, etc.)?

[MellowDrama]

11 hours ago, Lab Student said:

What is this tool actually for? conversion for what purposes?

Does it just take text and then make it into an mp3?

it is just an easy all in one.

Previously i was copying each line from se edit to eleven labs, downloading that in wav which was 3 extra button presses, the generating a lip file from blablabla

it was all rather tedious so this just does it all in one using Qwen3 TTS because xvasynth is medieval compared to what we have access to now and my ears can't take the abuse anymore.

 

2 hours ago, necol said:

Hi! First, thank you for the great tool.
I'm testing Qwen3-TTS-0.6B-CustomVoice with Skyrim Voice Tools.
English works perfectly.
However, when I set Language = Russian and use Cyrillic text, the generated speech is unintelligible (gibberish).
Is Russian currently supported by the backend/model? If yes, what language value should be used (Russian, ru, ru-RU, etc.)?

actually it never occurred to me to use anything else but English :p. The language setting is just for Qwen3 and only the English voice files are included.

I can fix this no problem though.

 

 

17 hours ago, bigbasi said:

Are you kidding me... I just spend the day starting to program my own tool, trying to create better results than with xvasynth 😉

 

congrats, for beating me to it 😉

 

Installer ran fine without errors, Path is configured

 

Anyway, I just ran into an error:

 

MD's Skyrim Voice Tools path is not configured in ExportDialogueVoiceJobsAndLaunchTools.pas.
Run the full installer, or edit VoiceGeneratorPath near the top of this script.

 

const
  {
    ...

  }
  VoiceGeneratorPath = 'E:\Nefaram 17.0.5\Tools\MD''s Skyrim Voice Tools\skyrim-audio-tools';

 

function ResolvedVoiceGeneratorPath: string;
begin
  Result := WithoutTrailingSlash(StripQuotes(VoiceGeneratorPath));

  if Result = 'E:\Nefaram 17.0.5\Tools\MD''s Skyrim Voice Tools\skyrim-audio-tools' then
    Result := '';
end;

I reran the installer and ran into the same issue. I'll be back home tonight to test but it shouldn't take long to fix.

 

 

20 hours ago, Pavlovian Afterdark said:

The Cicero sample line made me chuckle. Thank you for sharing.

If we're gonna be on LL anyway, might as well make the most of it.

[Courier07]

Can I use xVAsynth voice model files with this? Also can I use pre generated voice lines to generate new ones? Like, take a pre existing line, feed it to the generator, and tell it to make a new voice file for a new voice with the same pronunciation and cadence? Clone the voice lines, basically speaking.

[MellowDrama]

8 hours ago, Courier07 said:

Can I use xVAsynth voice model files with this? Also can I use pre generated voice lines to generate new ones? Like, take a pre existing line, feed it to the generator, and tell it to make a new voice file for a new voice with the same pronunciation and cadence? Clone the voice lines, basically speaking.

You could 

all you need is 30 seconds of that new voice

add it to the folder archive_recordings_trimmed

there's a text file called skyrim_voice_types.txt 
just add it at the bottom there and that should do the trick.

I'll try adding proper support for this later.

it's a good idea.

 

8 hours ago, Lab Student said:

Is this just a crossword puzzle app?
How does it work in games?

its not a mod that adds something to the game. It's a modder's resource type tool.

 

[MellowDrama]

run the .pas in se edit on the dialogue topic/ esp or write out you r lines manually in multiline.txt something something but i dont recommend that obviously.

Edited July 29 by MellowDrama

[bigbasi]

On 7/28/2026 at 4:17 PM, MellowDrama said:

I reran the installer and ran into the same issue. I'll be back home tonight to test but it shouldn't take long to fix.

 

I removed the check 
  if Result = 'E:\Nefaram 17.0.5\tools\MD''s Skyrim Voice Tools\skyrim-audio-tools' then
    Result := '';
since its redundant, now it works.

 

but now it cant find sox, which is installed and tries to load  the model into the cpu.... argh .... I'll try again 😉

[MellowDrama]

38 minutes ago, bigbasi said:

I removed the check 
  if Result = 'E:\Nefaram 17.0.5\tools\MD''s Skyrim Voice Tools\skyrim-audio-tools' then
    Result := '';
since its redundant, now it works.

 

but now it cant find sox, which is installed and tries to load  the model into the cpu.... argh .... I'll try again 😉

im almost done adding the other languages and a fix for that nonsense
and custom voices

gimme an hour or two 😛

[bigbasi]

29 minutes ago, MellowDrama said:

im almost done adding the other languages and a fix for that nonsense
and custom voices

gimme an hour or two 😛

Yeah, take your time... not doin' anything today anymore 😉 btw... I almost (99%) made it work with my approach... but xtts_v2 sometimes adds chinese to the output wav - but it's random, one time it does it ... the other time not... I am going crazy... so yeah - had enough for today 🤮 ... maybe I skip my Idea and switch to qwen

Edited July 29 by bigbasi

[MellowDrama]

19 hours ago, Courier07 said:

Can I use xVAsynth voice model files with this? Also can I use pre generated voice lines to generate new ones? Like, take a pre existing line, feed it to the generator, and tell it to make a new voice file for a new voice with the same pronunciation and cadence? Clone the voice lines, basically speaking.

done lemme know how it works out for you

 

On 7/28/2026 at 1:29 PM, necol said:

Hi! First, thank you for the great tool.
I'm testing Qwen3-TTS-0.6B-CustomVoice with Skyrim Voice Tools.
English works perfectly.
However, when I set Language = Russian and use Cyrillic text, the generated speech is unintelligible (gibberish).
Is Russian currently supported by the backend/model? If yes, what language value should be used (Russian, ru, ru-RU, etc.)?

added rusky

i dont speak it so lemme know if it works or not

[Courier07]

3 hours ago, MellowDrama said:

done lemme know how it works out for you

 

added rusky

i dont speak it so lemme know if it works or not

Hold on how do I use this ? I am sorry I am dumb as a rock when it comes to these audio related stuff. Where do I drop the old samples? Is there like a mini tutorial or something?

[necol]

Qwen3 TTS generating mzinbathedialogbathedirtytopic_02000058_1 (1/1) try 1/1 - 0%
Saving generated MP3...
Building FUZ: mzinbathedialogbathedirtytopic_02000058_1 (1/1)
Converting MP3 to WAV...
Writing LIP...
LipGenerator.exe, Version 1.0.0.2
FxAnalyze FAILED: The specified language is not supported.
Done. Exit code: 1 

It appears that LipGenerator.exe does not support the Russian language. Could you please look into this issue?

[bigbasi]

Got it to work, thanks again for sharing, but the Khajit voices are broken ... might be because you just used the shouts for training 😉

 

other than that, great work! the quality is really amazing, I started remaking all my voice packs!

Edited July 30 by bigbasi

[MellowDrama]

9 hours ago, necol said:

Qwen3 TTS generating mzinbathedialogbathedirtytopic_02000058_1 (1/1) try 1/1 - 0%
Saving generated MP3...
Building FUZ: mzinbathedialogbathedirtytopic_02000058_1 (1/1)
Converting MP3 to WAV...
Writing LIP...
LipGenerator.exe, Version 1.0.0.2
FxAnalyze FAILED: The specified language is not supported.
Done. Exit code: 1 

It appears that LipGenerator.exe does not support the Russian language. Could you please look into this issue?

Can you check again if it works properly for you now?

 

 

3 hours ago, bigbasi said:

Got it to work, thanks again for sharing, but the Khajit voices are broken ... might be because you just used the shouts for training 😉

 

other than that, great work! the quality is really amazing, I started remaking all my voice packs!

thanks thats weird.

 

Also I do suggest using this multiple samples 

That way you can catch errors before putting it into the game or opening of folders and playing mp3s

 

Khajits are fixed now.

 

Edited July 30 by MellowDrama

[bigbasi]

1 hour ago, MellowDrama said:

Khajits are fixed now.

 

thanks Buddy!

 

but I noticed that the Machine transcript is missing like in the other models... took the example from another. I don't know if the transcript is necceary for qwen
 

Machine transcript:
Model: small.en
Audio: femalecoward.mp3

I don't... something's happened. Blood of the sick, blood of the weak, blood of the unsuspecting noble lord. Savor each flavor. I keep an eye on their comings and goings. No one suspects poor Kireen, poor... the darkness. Where... need... What's happening? Who are you? Please, let me go! Why me? Ugh... Ugh... Huh? Help! My love! You came for me!

Timed segments:
[000.00-002.16] I don't... something's happened. avg_logprob=-0.37 no_speech_prob=0.50
..... and so on

Edited July 30 by bigbasi

[necol]

Yes, the files are now being packaged correctly into the FUZ format. However, the output audio is still gibberish instead of Russian speech. Oh well, thanks anyway for the tool. The quality of the English voiceover is very good.

[muravu]

Nice! Some time ago i tried to make something similar, had a very clunky thing and this seems to be much better. 

[MellowDrama]

53 minutes ago, necol said:

Yes, the files are now being packaged correctly into the FUZ format. However, the output audio is still gibberish instead of Russian speech. Oh well, thanks anyway for the tool. The quality of the English voiceover is very good.

all Russian sounds like gibberish to me so you know I tried my best. Can you tell me if it's all Russian voices or just specific ones? And is the Russian example i put in the description also gibberish?
Would you mind taking a look at the Russian sample folder? It's a bunch of mp3 s from the base game.

Maybe that's the issue.

[StuffedNinga]

It works great but it seems to want to convert each generated wav file to fuz immediately, which seems to slow down the generation process.  Each audio file takes about 30 seconds to complete a fuz file whereas other TTS generators take only about 4 seconds for me to complete a wav file.  Would it be possible to only generate wav files and leave the fuz generation process for the end, as a separate process, after all wav files have finished generating?  This would also help folks listen to individual audio files for issues and easy editing with Audacity before finalizing them as a fuz.

 

You also might not want to make it too easy to generate audio to completion as this completely bypasses quality control.  Mass generated audio will sound mostly good but you'd still need a human to listen to the generated files for errors and proper emotional context before finalizing it.  A one click production, like you currently have, is a breeze to run, but it also means folks are less likely to check the generated audio to make sure it's actually good.

Edited July 30 by StuffedNinga

[Courier07]

On 7/30/2026 at 6:37 AM, Courier07 said:

Hold on how do I use this ? I am sorry I am dumb as a rock when it comes to these audio related stuff. Where do I drop the old samples? Is there like a mini tutorial or something?

Any chance I could get some help or am I figuring this out myself? I understand nearly everything else, the only thing I don't get is the jobs field. It points to some example txt file, and so what am I supposed to do with that?

Edited July 31 by Courier07

[bigbasi]

17 hours ago, bigbasi said:

but I noticed that the Machine transcript is missing like in the other models

You dont need it! works fine without!

 

 

[Xarcellmo]

On 7/29/2026 at 7:30 PM, bigbasi said:

I removed the check 
  if Result = 'E:\Nefaram 17.0.5\tools\MD''s Skyrim Voice Tools\skyrim-audio-tools' then
    Result := '';
since its redundant, now it works.

 

but now it cant find sox, which is installed and tries to load  the model into the cpu.... argh .... I'll try again 😉

I got the same problem with sox.

[bigbasi]

7 minutes ago, Xarcellmo said:

I got the same problem with sox.

I install sox from the website, added the path to the executable to the windows path... and of course windows needed a reboot to make the settings work with MD's tool