diff --git a/README.md b/README.md index 731f532..561d42a 100644 --- a/README.md +++ b/README.md @@ -82,7 +82,7 @@ FFmpeg must be installed and configured for the application to process any track - The new auto-generated folder will be named after the model(s) selected. - The output audio files will be saved to the auto-generated directory. - The filenames for the instrumental & vocal outputs will have the selected model(s) name(s) appended. -- **Save All Outputs** - Only selectable when using the "*Ensemble Mode*" conversion method. This option will save all of the individual conversion outputs from each model within the ensemble +- **Save All Outputs** - Only selectable when using the "*Ensemble Mode*" conversion method. This option will save all of the individual conversion outputs from each model within the ensemble. - When *'Save All Outputs'* is un-selected, the application will auto-delete all of the individual conversions generated by each model in the ensemble. ### Additional Options @@ -92,13 +92,11 @@ FFmpeg must be installed and configured for the application to process any track - **1024** - Low conversion quality, shortest conversion time, low resource usage - **512** - Average conversion quality, average conversion time, normal resource usage - **320** - Better conversion quality, long conversion time, high resource usage - - **Aggression Setting** - This option allows you to set how strong the vocal removal will be. - The range is 0-100. - Higher values perform deeper extractions. - The default is 10 for instrumental & vocal models. - Values over 10 can result in muddy-sounding instrumentals for the non-vocal models. - - **Default Values:** - **Window Size** - 512 - **Aggression Setting** - 10 (optimal setting for all conversions) @@ -109,16 +107,14 @@ FFmpeg must be installed and configured for the application to process any track ## Models Included -All of the models included in the release were trained on large datasets containing a diverse range of music genres as well as different training parameters. +All of the models included in the release were trained on large datasets containing a diverse range of music genres and different training parameters. -- **PLEASE NOTE:** Do not change the name of the models provided! The required parameters are specified and appended to the end of the filenames. +**Please Note:** Do not change the name of the models provided! The required parameters are specified and appended to the end of the filenames. - **Model Network Types** - **HP2** - The model layers are much larger. However, this makes them resource-heavy. - **HP** - The model layers are the standard size for UVR v5. -Here's a list of the models included within the package - - ### Main Models - **HP2_3BAND_44100_MSB2.pth** - This is a strong instrumental model trained using more data and new parameters. @@ -137,36 +133,40 @@ Here's a list of the models included within the package - Run your tracks through a single model only. This is the default conversion method. - **Choose Main Model** - Here is where you choose the main model to perform a deep vocal removal. - - The *'Model Test Mode'* option makes it easier for the user to test different models on given tracks. + - The *'Model Test Mode'* option makes it easier for users to test different models on given tracks. ### Ensemble Mode -Ensemble Mode will run your track(s) through multiple models and combine the resulting outputs for a more robust separation. +Ensemble Mode will run your track(s) through multiple models and combine the resulting outputs for a more robust separation. Higher level ensembles will have stronger separations, as they use more models. -- **Choose Ensemble** - Here, you choose the ensemble you wish to run your track through. - - The *'Save All Output'* option saves all of the outputs generated by each model in the ensemble when checked. - -Currently, there are 4 ensembles you can choose from. +- **Choose Ensemble** - Here, choose the ensemble you wish to run your track through. + - **There are 4 ensembles you can choose from:** + - **HP1 Models** - Level 1 Ensemble + - **HP2 Models** - Level 2 Ensemble + - **All HP Models** - Level 3 Ensemble + - **Vocal Models** - Level 1 Vocal Ensemble + - A directory is auto-generated with the name of the ensemble. This directory will contain all of the individual outputs generated by the ensemble and auto-delete once the conversions are complete if the *'Save All Outputs'* option is unchecked. + - When checked, the *'Save All Outputs'* option saves all of the outputs generated by each model in the ensemble. - **List of models included in each ensemble:** - - **HP1 Models** - Level 1 Ensemble + - **HP1 Models** - HP_4BAND_44100_A - HP_4BAND_44100_B - - **HP2 Models** - Level 2 Ensemble + - **HP2 Models** - HP2_4BAND_44100_1 - HP2_4BAND_44100_2 - HP2_3BAND_44100_MSB2 - - **All HP Models** - Level 3 Ensemble + - **All HP Models** - HP_4BAND_44100_A - HP_4BAND_44100_B - HP2_4BAND_44100_1 - HP2_4BAND_44100_2 - HP2_3BAND_44100_MSB2 - - **Vocal Models** - Level 1 Vocal Ensemble + - **Vocal Models** - HP_Vocal_4BAND_44100 - HP_Vocal_AGG_4BAND_44100 -- **PLEASE NOTE:** Ensemble mode is very resource heavy! +- **Please Note:** Ensemble mode is very resource heavy! ## Other GUI Notes @@ -201,7 +201,7 @@ python VocalRemover.py The **Ultimate Vocal Remover GUI** code is [MIT-licensed](LICENSE). -- **PLEASE NOTE:** For all third-party application developers who wish to use our models, please honor the MIT license by providing credit to UVR and its developers Anjok07, aufr33, & tsurumeso. +- **Please Note:** For all third-party application developers who wish to use our models, please honor the MIT license by providing credit to UVR and its developers Anjok07, aufr33, & tsurumeso. ## Contributing