- Tools to create a dataset for your own trained voice model with your own FFXIV GameData.
- FOR PERSONAL USE ONLY.
- Using this tool I expect you to know how to use alltalk_tts. If you need any assistance feel free to
- This tool is at the moment only tested in german but should still work in the other official client languages(english, french, japanese)
- Download latest release
- Extract to a place of your liking
- Make sure you got .NET Framework Core 8.0 installed: .Net Core download
- Start "Echokraut-Tools.exe"
- Select your game install location (needs to contain "game" and "boot" folder), defaults to the default install location of FFXIV (without steam).
- Select save location for the extracted data (should have atleast 10 to 20GB of free space), defaults to the location where my Echokraut Plugin installs alltalk.
- Select your language
- Hit the button.
- After it loads all necessary exd data it will begin extracting the audio files and converting them to a usable format. (A separate CMD window will open for that process, don't do anything and let it work.)
- The progress bar will show "done" when it's done. You can now close "Echokraut Tools"
- Copy "metadata_train.csv", "metadata_eval.csv" and the "wavs" folder into a new folder (FFXIV for example) inside the "finetune" folder of alltalk_tts.
- Start the finetune process according to the alltalk manual.
- Once the finetune process is finished: You can create your voice files by taking one of the many .wav files out of the wavs/npcname folders for each npc and test them in alltalk_tts until you find one you are happy with. And name them according to the Echokraut-Plugin instructions.
