Intunist Windows ETK v2.0.0-wip1

Download "Intunist-Windows-ETK-2.0-wip1.rar" for the local Windows ETK.
This release is completely pre-configured and nothing more needs to be downloaded after.

RELEASE UPDATE: Please run "INSTALL PYTORCH.bat" first before using. The current version is too large to include even when compressed.

• Update Python from 3.8.10 > 3.8.16, from scratch.
• Update all packages to latest versions.
• Update feature generation configuration.
• Correct mistakes in "Change Settings" file.
• Update HED file to latest intunist_jp version.

This is ready to train japanese as-is. You do not need to change any configuration if you do not want or need to.

PyTorch and other required packages are pre-installed to avoid reoccurring issues. Please edit "change_settings.bat" to choose the model type you would like to train.
By default it is ResSkipF0FFConvLSTM.

Intunist Windows ETK v1.0.3.1

a mistake was found in the change-settings.sh file. this has been corrected

Intunist Windows ETK v1.0.3

v1.0.3 update: fix change_settings.sh, fix install script.
Updated ENUNU Training Kit for training locally on Windows. Based on current NNSVS. All packages are updated to latest version.
Based on Portable ENUNU Training Kit.
Useful for feature generation before training on Colab or GPU rental service.

1. Edit and run change_settings.bat
   • Instructions are inside file
2. Edit acoustic_current.yaml if desired
   • Required if you have less VRAM, lower hidden_dim value as required
3. Place dataset in singing_database
4. run "RUN-TO-TRAIN" bat file to train model, pick version based on what was chosen in "change_settings.bat"

By default only supports Intunist Japanese+.
For ResF0 you will need to edit:
\conf\train\acoustic\model\acoustic_current.yaml, \conf\train\acoustic\data\myconfig.yaml for the correct idx settings for your hed.
For any non-japanese hed you will need to edit:
\stage0\compare_mono_align_and_mono_score.py and update VOWELS list.

Intunist Windows ETK v1.0.2

v1.0.2 update: fix change_settings.sh
Updated ENUNU Training Kit for training locally on Windows. Based on current NNSVS. All packages are updated to latest version.
Based on Portable ENUNU Training Kit.
Useful for feature generation before training on Colab or GPU rental service.

1. Edit and run change_settings.bat
   • Instructions are inside file
2. Edit acoustic_current.yaml if desired
   • Required if you have less VRAM, lower hidden_dim value as required
3. Place dataset in singing_database
4. run "RUN-TO-TRAIN" bat file to train model, pick version based on what was chosen in "change_settings.bat"

By default only supports Intunist Japanese+.
For ResF0 you will need to edit:
\conf\train\acoustic\model\acoustic_current.yaml, \conf\train\acoustic\data\myconfig.yaml for the correct idx settings for your hed.
For any non-japanese hed you will need to edit:
\stage0\compare_mono_align_and_mono_score.py and update VOWELS list.

Intunist Windows ETK v1.0.1

v1.0.1 update: minor correction
Updated ENUNU Training Kit for training locally on Windows. Based on current NNSVS. All packages are updated to latest version.
Based on Portable ENUNU Training Kit.
Useful for feature generation before training on Colab or rental service.

1. Edit and run change_settings.bat
   • Instructions are inside file
2. Edit acoustic_current.yaml if desired
   • Required if you have less VRAM, lower hidden_dim value as required
3. Place dataset in singing_database
4. run "RUN-TO-TRAIN" bat file to train model, pick version based on what was chosen in "change_settings.bat"

Intunist ETK Windows v1.0.0

Updated ETK for training locally on Windows. Based on current NNSVS. All packages are updated to latest version.
Based on Portable ENUNU Training Kit.

1. Edit and run change_settings.bat
2. Edit acoustic_current.yaml if desired
3. Place dataset in singing_database
4. run run_resf0.bat or run_non-resf0.bat
5. run release.bat