2 new models and minor fix to README #165
Merged
+8
−1
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Nice one, thank you so much for raising this PR and for making it super easy to action! I'm just waiting for the model checkpoint files to upload to this repo's models bucket but I've just tested on my mac and linux machines. I'll merge and release a new |
|
Released in |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Models:
Mel Band Roformer Vocal by Kimberley Jensen
SDR Vocals: 11.01
SDR Instrumental: 17.32
Checkpoint Link: ckpt
Config Link: config
The original config file doesn't work, so I found one that does work.
MelBand Roformer Kim | Inst V1 (E) by Unwa
SDR Instrumental: 16.36
SDR Other: 10.05
Checkpoint Link: ckpt
Config Link: config yep same file as MelBand Roformer Kim | Inst V1 by Unwa
Note: Max instrum fullness, but a bit noisy
(E) means Emphasis, the e models by unwa emphasize the fullness metric and thus have the highest score in that metric.
Both models were tested on Windows 11 and Ubuntu 24.04, but only with CPU as I don't have a GPU. I hope they work well with GPU and on MacOS.
README fix:
Just a minor fix to the default amplification_threshold value, the default value is now 0.0.