New De-reverb model
A new MDX23C De-reverb model has been added.
Weekly report
1. In the new interface, the BS-RoFormer model now also has De-mudder.
2. We have started work on two new models: RoFormer De-reverb and De-echo. We are also experimenting with a new model for backing vocals isolation.
New user interface
The new UI is now available for testing!
2024-07-15 06:09:02 +0000 UTC
View Post
Which version is better
We're going to redesign the UI a little bit. The current version is pretty complex for many users.
We're considering two versions.
Please vote!
1. Remove the quick presets. You just select the Separation type.
Weekly report
1. Added new Demudder post-processing for Mel-Roformer model. It perfectly eliminates the distortion that occurs when vocals are removed. Now you can forget about "max_mag (drums)"!
For best results, we recommend using songs in the highest quality WAV or FLAC format.
Two new premium models
New models have been added:
1. BS-RoFormer Bass - a new model for bass isolation, which in many cases is better than Demucs.
2024-05-23 10:20:24 +0000 UTC
Mel-Roformer Piano beta model
A new model for piano isolation has been added. This is our very first model and it is not perfect yet, but it works very well with some songs.
You can find it in the menu or by clicking on the "Piano" shortcut.
2024-05-21 04:47:50 +0000 UTC
Mel-Roformer Karaoke model
Added a new Mel-Roformer model for lead vocal isolation.
Compared to the previous ones it produces better results on some songs.
2024-05-16 10:01:50 +0000 UTC
Weekly report
1. A new max_mag (drums) ensemble algorithm has been added for the Mel-RoFormer model. It allows you to enhance the drums in some songs, making them less muddy.
2. Added a new crowd removal model.
2024-05-10 02:42:22 +0000 UTC
Added max_mag ensemble for Mel-RoFormer model
It combines Mel-RoFormer and BS-RoFormer results, making the instrumentals even less ...
2024-04-29 11:20:36 +0000 UTC
View Post
New Mel-Roformer model
A new model has been added to remove vocals. Unlike the BS-Roformer, the new model produces less muddy instrumentals.
This model should not be used on songs with sax, duduk or similar instruments. But in many cases it is the best model so far.
2024-04-25 07:21:48 +0000 UTC
Weekly report
1. Added a new setting for the BVE models.
You can now explicitly specify the content type of the audio file:
🞄 auto - automatic detection based on file name.
🞄 full mix - the MDX model will be used for preprocessing. In this mode, you will get all stems. Previously, this option was the default.
🞄 vocals only - MDX model will NOT be use...
2024-04-21 07:56:02 +0000 UTC
View Post
Weekly report
1. Added mixing of MDX23C and BS-RoFormer results.
So far it works only for MDX23C.
2024-04-07 05:27:23 +0000 UTC
New vocal isolation model
BS-RoFormer has just been added to the X-Minus / UVR Online!
This is a HQ model for music and vocal separation.
Many thanks to viperx for model training.
2024-03-29 00:24:16 +0000 UTC
De-noise model update
A new version of the UVR De-noise model has been added. This model is designed for low noise removal. Unlike the previous version, it is less aggressive and removes only uniform noise, preserving SFX and all instruments in the music.
Use the following settings:
AI model - UVR De-noise
Aggressiveness - minimum
<...
2024-03-23 07:28:05 +0000 UTC
View Post
Information for users who are unable to pay on Boosty
We've just restored the Patreon page!
If your last payment through Boosty failed, please subscribe here:
Weekly report
New models continue to be developed. Here's what's expected in the near future:
1. A less aggressive version of the UVR De-noise model. It is designed to remove small uniform noise. This model is expected late this week or early next week.
2. Piano model. It will most likely be ready in early May.
3. A new MDX model for vocal removal.
Weekly report
1. Added QR code download button
Weekly report
This week was a very busy week, but we added two new features:
1. Lead vocal download button after processing by the UVR BVE model
2. volume control (at the top of the page)
2024-02-24 04:57:13 +0000 UTC
New MDX De-crowd model
Another crowd removal model has been added. This is an alternative to UVR De-crowd.
De-crowd model
A new UVR De-crowd beta model has been added. This model is designed to remove crowd and applause. It is located at the bottom of the menu.
2024-01-29 00:48:54 +0000 UTC
Two little news
1. The De-crowd model is almost ready to start training. We'll add the UVR version first, then the MDX version. UVR De-crowd is expected to be ready at the end of this week.
2. Rebranding. We changed the name of the service from X-Minus.pro to UVR Online. So far, only the logo has been replaced. The website URL remains the same.
Weekly report
1. The creation of the De-crowd / De-applause model is ongoing. It will be ready sometime in late January or early February.
2. Small changes in the UI. Now there are more quick presets and they are located at the top so that new users can easily select the desired model.
New noise reduction model for speech
This model only works with speech, not music. For music, please use UVR De-noise model.
Happy New Year!
We just added a Demucs drumsep model to split the drums into components: cymbals, kick, snare and toms. This is the last model in 2023.
2023-12-30 05:50:05 +0000 UTC
Merry Christmas! 🎅 🎄
We already have 7,000 subscribers. Thank you for being with us!
We will start next year with new models and some new features. We're going to be adding these models:
1. A new De-noise model, instead of the previously announced MDX De-noise. It is not intended for music, but it cleans the speech even better than our last UVR De-noise model.
UVR De-noise model
We just added the UVR De-noise beta model!
UPD: The model is ready.
The latest news about the MDX De-noise model
As we posted earlier, due to technical reasons this model is delayed.
However, we have two good news:
1. A similar model UVR De-noise by FoxJoy will be added tomorrow. It is not designed to remove heavy noise, but it may still be useful for someone.
UPD: this model has been added!
2. We thought it wou...
2023-12-09 10:48:28 +0000 UTC
View Post
Weekly report
1. De-noise model training has started! We'll have the first results by mid-December (or the end of December).
2. The number of minutes on Ultimate level has been increased up to 600.
3. We've updated the model information. To read it, click on the little "?" button.
2023-12-03 01:40:49 +0000 UTC
