================================ Please note: ================================ all generation tools will use your saved browser settings for your elevenlabs account
Log Saving/Loading is done with a csv file, matching the batch generation format. You can also load a csv to the log to view it, if desired.
================================ TTS batch Generation ================================ Requires A CSV file, you can generate one with dialogue from Xedit Use google sheets or Excel or similar app to move columns around
CSV columns one and two should match this format: Column 1, Column 2 "Filename","Response Text"
Quotations are only needed if the text contains special characters or punctuation.
================================ CSV File Copying Help ================================ Note; CSV columns 3 & 4 are used for batch file copying.
You can add tags to these columns so that when you can copy files they will be added to a new subdirectory.
Ex: Filename1,"I'm your wife" , "Lite Voices,Female only" , "" Filename2 , "I don't believe you" , "" , ""
Would output 3 subfolders in the target directory: "Lite Voices", "Female Only", and "Unspecified"
"I'm your wife" would be copied to both subfolders "lite voices" and "Female Only" while "I don't believe you" would be copied to the "unspecified" subfolder.
================================ Batch File Renaming ================================ Rename files based on a CSV file You can use this to convert filenames to the DBVO format
Note: Filenames that contain unsupported characters will have those characters converted to "_" and files that are duplicates will have "(#)" appended to the filename.
================================ STS Batch Generation ================================ Can use wav or mp3 files
All files from the target directory will be iterated through
If the output directory contains a file with the same name as the input directory The file will be skipped.
Must use a Language model that supports "STS"
================================ File output for single generations ================================ If no filename is specified the file will simply be saved with a timestamp for its filename
If no Output directory is specified A folder will be created in the Exe's location to store the files
For STS regeneration A "Recordings" folder will be created in the output directory With sent audio stored there for log regeneration Delete at your discretion, but know the app has no way to get this info back
I'm still hoping you'll release an update fixing the memory leak, and adding Output Path and Hotkey to the INI, plus storing the dialogue string in the Title field of the meta data.
If you have a csv that seperates NPCs out by race you can use google sheets to sort it based on the column that has the Race flag. Then seperate those out into individual csv files for each race.
My tool doesn't make use of any distinctions like that. You could place the race into the 3rd or 4th columns and use the copying feature to seperate them out, but you would still be generating all the voicelines.
Thank you for the v3 release, the memory leak definitely appears to be fixed, I generated 559 lines of dialogue without closing and relaunching (v2 failed shortly after 200 lines).
There's still an obscure issue when naming the first file generated from an Excel-created CSV file. You've resolved the prefixing of "" but the filename is still coming out wrong.
Within Windows itself, the filename appears fine. The Oblivion Construction Set cannot see the MP3 file converted from the WAV file. Attempting to copy/paste the file in the same directory throws an error but a copy is created. It's only when using a DOS prompt and running the DIR command that I discovered this; viewed in the DOS prompt, it appears to have a space added after the first character of the filename (not visible in Windows), but if I save the DIR result into a text file, it instead displays as a question mark prefixing the filename.
Windows: AGTopics_AgJoin_0000E414_1.wav
DOS prompt using DIR (may not successfully paste here): 26/02/2025 15:58 601,388 AGTopics_AgJoin_0000E414_1.wav
Text file generated using DIR > t.txt: 26/02/2025 15:58 601,388 ?AGTopics_AgJoin_0000E414_1.wav
Renaming the file in Windows appears to work, but if I then revert the filename back again, it restores this hidden character once more. I was able to use the DOS REN command to successfully rename it, after which Windows and the Oblivion Construction Set treat the file normally at last.
EDIT Really weird, now Windows renaming is working where previously it wasn't. Please see if you could fix this, it's harder to spot, at least in v2 the misnamed file was clearly identifiable, now the additional character is hidden so easily forgotten.
Is the problem with the first line exclusive to Excel spreadsheets or does this occur with google sheets as well? I have not come across this issue.
What program are you using to convert the filetype? If you are only renaming the file, that wouldn't change its encapsulation, only how windows reads it (maybe some file headers but I'm not an expert on that). Perhaps try a free program like VLC, that has a batch conversion function.
I use Yakitori Audio Convertor to batch convert, it had no issues so clearly is happy to repeat the hidden character in the filename. I'll be doing some more generating tomorrow, so will try creating a CSV from a txt file to experiment. You may recall that when I first reported the filename issue for v2, there was no problem from txt files saved as CSV, but those require a lot of work to manually compile whereas Excel handles the layout automatically.
It looks like excel might be saving the CSV files to UTF-8 with BOM (Byte order marker)? This could cause problems now that the program reads with UTF-8 as that special character is now added to the filename.
Purely conjecture at this point though, I still have to do a lot more testing and I don't have excel personally.
I've found the option in Excel to select the encoding. It defaults to "Western European (Windows)" not UTF-8, though it can be changed to "Unicode (UTF-8)". In Excel, excluding the options for other parts of the world, the options are: Unicode, Unicode (Big-Endian), Unicode (UTF-8), US-ASCII, Western European (ISO), Western European (Windows). My preferred text editor is EditPlus, which defaults encoding to "ANSI", but includes the options "UTF-8" and "UTF-8 + BOM". Excel does not include an option mentioning BOM.
I just used a CSV created with Google Sheets, and another created with EditPlus, no filename issues with either of those batches.
So excel must be adding the character for some reason then? Interesting. Ill look at the sheets you shared with me and maybe I can figure out whats going on.
Thank you, an excellent successor to Mathiew's DVBO Voice Pack Maker, especially as you support the Turbo models, meaning half the credits spent for the same number of characters.
Please could you add the response text into the metadata's Title field, so each WAV file contains a record of the words spoken?
That's a great idea, I'll definitely consider doing it. Working on sts batch generation too.
The app gets a list of models direectly from elevenlabs so it's futureproofed for new models as well, assuming elevenlabs keeps the same information format.
21 comments
================================
Please note:
================================
all generation tools will use your saved browser settings for your elevenlabs account
Log Saving/Loading is done with a csv file, matching the batch generation format. You can also load a csv to the log to view it, if desired.
================================
TTS batch Generation
================================
Requires A CSV file, you can generate one with dialogue from Xedit
Use google sheets or Excel or similar app to move columns around
CSV columns one and two should match this format:
Column 1, Column 2
"Filename","Response Text"
Quotations are only needed if the text contains special characters or punctuation.
================================
CSV File Copying Help
================================
Note; CSV columns 3 & 4 are used for batch file copying.
You can add tags to these columns so that when you can copy files they will be added to a new subdirectory.
Ex:
Filename1,"I'm your wife" , "Lite Voices,Female only" , ""
Filename2 , "I don't believe you" , "" , ""
Would output 3 subfolders in the target directory: "Lite Voices", "Female Only", and "Unspecified"
"I'm your wife" would be copied to both subfolders "lite voices" and "Female Only" while "I don't believe you" would be copied to the "unspecified" subfolder.
================================
Batch File Renaming
================================
Rename files based on a CSV file
You can use this to convert filenames to the DBVO format
Note: Filenames that contain unsupported characters will have those characters converted to "_" and files that are duplicates will have "(#)" appended to the filename.
================================
STS Batch Generation
================================
Can use wav or mp3 files
All files from the target directory will be iterated through
If the output directory contains a file with the same name as the input directory The file will be skipped.
Must use a Language model that supports "STS"
================================
File output for single generations
================================
If no filename is specified
the file will simply be saved with a timestamp for its filename
If no Output directory is specified
A folder will be created in the Exe's location to store the files
For STS regeneration
A "Recordings" folder will be created in the output directory
With sent audio stored there for log regeneration
Delete at your discretion, but know the app has no way to get this info back
Black Horse Courier Expanded Voices (ElevenLabs)
Dark Brotherhood Continued Voices (ElevenLabs)
Dibellas Watch Voices (ElevenLabs)
Enhanced Economy Voices (ElevenLabs) (only voice files added in v4, the rest were generated using DBVO Voice Pack Maker before I discovered ElevenLabs Batch Generator)
Fighters Guild Quests Voices (ElevenLabs)
Hoarfrost Castle Voices (ElevenLabs)
Mazoga Follows Voices (ElevenLabs)
Elsweyr the Deserts of Anequina Voices (ElevenLabs)
Stirk Voices (ElevenLabs)
The Return of the Dark Brotherhood Voices (ElevenLabs)
Vampire Hunting - Order of the Virtuous Blood Voices (ElevenLabs)
I also generated voice files using ElevenLabs Batch Generator for Firefly101 to include in his mod Firefly101 Mercenaries
Advanced Follower Framework Voices (ElevenLabs)
AFK Weye Voices (ElevenLabs)
CM Partners Mod Basic Voices (ElevenLabs)
Cyrodiil Transportation Network 2.0 Voices (ElevenLabs)
Gamblers of Cyrodiil Voices (ElevenLabs)
Integration - The Stranded Light Integrated Voices (ElevenLabs)
Kragenir's Death Quest Voices (ElevenLabs)
Mannimarco Resurrection Voices (ElevenLabs)
Mods Aware M'aiq Voices (ElevenLabs)
MTC Expanded Villages Voices (ElevenLabs)
Oblivion Adventurers Guild Voices (ElevenLabs)
Oblivion Marriage Voices (ElevenLabs)
Order of the Dragon EV Voices (ElevenLabs)
rhk DLC Delayed Voices (ElevenLabs)
SI Cities Expanded Voices (ElevenLabs)
Tales of Cyrodiil Voices (ElevenLabs)
Tales from Elsweyr Anequina-Rev's Redo Voices (ElevenLabs)
The Duelists Voices (ElevenLabs)
The Golden Crest Voices (ElevenLabs)
The Imperial Real Estate Voices (ElevenLabs)
The Vampires of Fanesburrow Voices (ElevenLabs)
Valenwood Improved Voices (ElevenLabs)
Woodland Rangers Revised Voices (ElevenLabs)
I have proper pas scripts that do this well for XVA but idk how it will work here
My tool doesn't make use of any distinctions like that. You could place the race into the 3rd or 4th columns and use the copying feature to seperate them out, but you would still be generating all the voicelines.
There's still an obscure issue when naming the first file generated from an Excel-created CSV file. You've resolved the prefixing of "" but the filename is still coming out wrong.
Within Windows itself, the filename appears fine. The Oblivion Construction Set cannot see the MP3 file converted from the WAV file. Attempting to copy/paste the file in the same directory throws an error but a copy is created. It's only when using a DOS prompt and running the DIR command that I discovered this; viewed in the DOS prompt, it appears to have a space added after the first character of the filename (not visible in Windows), but if I save the DIR result into a text file, it instead displays as a question mark prefixing the filename.
Windows:
AGTopics_AgJoin_0000E414_1.wav
DOS prompt using DIR (may not successfully paste here):
26/02/2025 15:58 601,388 AGTopics_AgJoin_0000E414_1.wav
Text file generated using DIR > t.txt:
26/02/2025 15:58 601,388 ?AGTopics_AgJoin_0000E414_1.wav
Renaming the file in Windows appears to work, but if I then revert the filename back again, it restores this hidden character once more. I was able to use the DOS REN command to successfully rename it, after which Windows and the Oblivion Construction Set treat the file normally at last.
EDIT
Really weird, now Windows renaming is working where previously it wasn't.
Please see if you could fix this, it's harder to spot, at least in v2 the misnamed file was clearly identifiable, now the additional character is hidden so easily forgotten.
What program are you using to convert the filetype? If you are only renaming the file, that wouldn't change its encapsulation, only how windows reads it (maybe some file headers but I'm not an expert on that). Perhaps try a free program like VLC, that has a batch conversion function.
Purely conjecture at this point though, I still have to do a lot more testing and I don't have excel personally.
In Excel, excluding the options for other parts of the world, the options are: Unicode, Unicode (Big-Endian), Unicode (UTF-8), US-ASCII,
Western European (ISO), Western European (Windows).
My preferred text editor is EditPlus, which defaults encoding to "ANSI", but includes the options "UTF-8" and "UTF-8 + BOM". Excel does not include an option mentioning BOM.
I just used a CSV created with Google Sheets, and another created with EditPlus, no filename issues with either of those batches.
Please could you add the response text into the metadata's Title field, so each WAV file contains a record of the words spoken?
The app gets a list of models direectly from elevenlabs so it's futureproofed for new models as well, assuming elevenlabs keeps the same information format.
- It works.
I created a bat file manually a long time ago using lazy voice finder. It was a pain in the butt lol. Really happy we have this tool now.
(Edited to sound less gloaty, sorry about that <3)