Cool_Tools/ultimatevocalremovergui
1
0
Fork 0
mirror of https://github.com/Anjok07/ultimatevocalremovergui.git synced 2024-11-24 07:20:10 +01:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
aufr33 2021-05-09 11:47:49 +03:00 committed by GitHub
parent 8244ad1617
commit 1d34200862
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
README.md
View File

@ -37,10 +37,11 @@ Please note, this version is based on vocal-remover 4.0.0 of tsurumeso's origina
- New format of spectrograms. Instead of a single spectrogram with a fixed FFT size, combined spectrograms are now used. This version combines several different types of spectrograms within specific frequency ranges. This approach allowed for a clearer view of the high frequencies and good resolutions at low frequencies, thus allowing for more targeted vocal removals.
- The arguments --sr, --n_fft, --hop_length are removed. JSON files are now used instead.
- The following new features were added
- **--high_end_process** - This argument restores the high frequencies of the instrumental (but not the vocals). The 3 choices for this argument are:
- **--high_end_process** - This argument restores the high frequencies of the instrumental (but not the vocals). It is intended for models with a narrow bandwidth, 16 kHz and below. The 4 choices for this argument are:
- *none* - No processing (default)
- *bypass* - This copies the missing frequencies from the input.
- *correlation* - This also copies missing frequencies from the input, however, the magnitude of the copied frequency will depend on the magnitude of the generated instrumental's high frequencies.
- *mirroring* - This is a more advanced algorithm than correlation. It uses the high frequencies from the input and mirrored instrumental's frequencies.
- **--aggressiveness** - This argument allows you to set how strong the vocal removal will be. The range is 0.0-1.0 The higher the value, the more the vocals will be removed. Please note, the highest value can result in muddy sounding instrumentals depending on the track being converted, so this isn't always recommended. The default is 0.1. For the vocal model specifically, the recommended value is 0.5-0.6.
- **--deepextraction** - This argument generates an additional instrumental output with deep artifact vocal removal. This option is experimental and is more suited for acoustic or other light types of tracks with stubborn vocals. Many others might sound bad.
@ -56,7 +57,7 @@ Here's a list of the models included within the v5 beta package -
- **2band_32000 Models**
- **MGM-v5-2Band-32000-BETA1.pth** - This model does very well on lower frequencies. Frequency cut-off is 16000 Hz. Must be used with **2band_32000.json** file!
- **MGM-v5-2Band-32000-BETA2.pth** - This model does very well on lower frequencies. Frequency cut-off is 16000 Hz. Must be used with **2band_32000.json** file!
- **MGM-v5-KAROKEE-32000-BETA1.pth** - Model by aufr33. This model focuses on removing main vocals only, leaving the BV vocals mostly intact. Frequency cut-off is 16000 kHz. Must be used with **2band_32000.json** file!
- **MGM-v5-KAROKEE-32000-BETA1.pth** - Model by aufr33. This model focuses on removing main vocals only, leaving the BV vocals mostly intact. Frequency cut-off is 16000 Hz. Must be used with **2band_32000.json** file!
- **MGM-v5-KAROKEE-32000-BETA2-AGR.pth** - Model by aufr33. This model focuses a bit more on removing vocals from lower frequencies.Frequency cut-off is 16000 Hz. Must be used with **2band_32000.json** file!
- **MGM-v5-Vocal_2Band-32000-BETA1.pth** - This is a model that provides cleaner vocal stems! Frequency cut-off is 16000 Hz. Must be used with **2band_32000.json** file!
- **MGM-v5-Vocal_2Band-32000-BETA2.pth** - This is a model that provides cleaner vocal stems! Frequency cut-off is 16000 Hz. Must be used with **2band_32000.json** file!