Weekly report
1. UVR BVE v2 model training is over.
2. We're going to start training the De-noise model soon.
3. Added MDX23C 8K model for quality vocal and music separation.
2023-11-18 04:58:27 +0000 UTC
BVE model update
The second version of the UVR BVE model is already available for testing!
A small update to the backing vocals extractor
Now you can more accurately specify the panning of the lead vocal.
Weekly report
1. Added support for stem.mp4 (Native Instruments). This format is supported by Traktor Pro and some other software. This only works with the Demucs V4 model so far.
2. When using the UVR BVE model, you can now download a song without lead vocals (instrumental with backing vocals).
3. Since we didn't get good results with Demucs BVE, we started training the se...
2023-10-29 06:08:25 +0000 UTC
View Post
New BVE stems
Added additional download buttons when using UVR BVE model.
Announcement of the new BVE model
The day before yesterday, we started training the M̶D̶X̶ Demucs BVE model (backing vocals extraction). The model will be ready sometime in the second half of November.
BVE model poll
We have an idea for the UVR BVE model. We're going to add two extra stems so that it becomes four:
- instrumental with lead vocals
- backing vocals
- instrumental without vocals
- all vocals
It won't consume audio minutes. If we implement it, will you use this feature?
Announcement of new models
We will be adding two more new models this fall:
1. BVE model for MDX
2. De-noise model to remove any kind of noise from voice or music
💡 We're still accepting your ideas for new models.
UVR BVE beta model
We now have the world's first Backing Vocal Extraction model!
This model is... 2023-08-08 05:36:31 +0000 UTC View Post
Announcement of a new BVE model
We are pleased to announce that the BVE (backing vocals extractor) model is almost complete. The estimated date for adding the model is August 9.
New MDX23C beta model
A new MDX23C beta model has been added!
The model is still in the training phase and will be improved in the coming weeks.
Please note that this model is very resource intensive so the maximum song length is limited to 15 minutes.
Announcement of a new MDX model
A new MDX23C model for vocal isolation will be added this week. This model is amazing. It has reached SDR 9.71 for vocals and over 16 for music!
This model is very resource intensive, so we will have to prepare the servers to run it. We expect to add it to the site this week.
New MDX Voc FT model
The new MDX Voc FT model is already available on X-Minus! This is one of the best models for extracting vocals.
The "mdx KimberleyJSN" model has been renamed to "mdx Kim inst".
The "mdx KimberleyJSN" model has been renamed to "mdx Kim inst".
New vocal model added
The mdx v2 (voc) model has been replaced by mdx Kim voc 2.
New mdx hq3 model
A new mdx hq3 model has been added. This model is an improved version of mdx full band.
2023-06-12 06:04:47 +0000 UTC
Weekly report (May 28)
1. Increased the maximum length of a song on Standard or higher subscription levels.
Weekly report (April 29)
1. Added a MDX De-reverb model. This model removes reverberation.
2. Added a UVR De-echo model. This model removes echo / delay.
Thanks to FoxJoy for providing the models!
3. The Backing Vocals Extractor dataset is 50% complete. We will start training the model around July.
...
2023-04-29 12:10:41 +0000 UTC
View Post
How to isolate backing vocals
While our B.V.E. model is under development, we want to tell you how you can extract/isolate the backing vocals now.
Method one
1. Remove the lead vocals from the song with these settings: AI model → mdx (kar) v2 or uvr (kar) v2. Download the instrumental.
2. Extract the backing vocals from the instrumental with the mdx v2 (voc) model or Dem... 2023-04-22 06:25:59 +0000 UTC View Post
Let's make a list of test songs
The development of the backing vocals (BVE) model is ongoing. And until it is ready, we ask that you help us make a list of test songs.
Please write in the comments the song titles or links to YouTube that we should focus on. These songs should contain a lot of backing vocals. We prioritize famous songs that have millions of listens.
We will try to optimize our model ...
2023-04-15 00:41:52 +0000 UTC
View Post
Premium (Lite+ / Lite 2024)
ENG: Thank you for your subscription!
Here's your secret link:
-----
РУС: Спасибо за вашу поддержку!
Для начала работы перейдите по ссылке:
March Update
1. The performance of the mdx models is improved. Processing is now faster.
2. The b.v. models have been renamed to kar.
3. A new Soprano voice setting has been added for songs with the high-pitched vocals.
This only works with mdx models so far.
MDX full band
A new mdx full band model has been added. Compared to the old versions, the separation quality is even higher and the bandwidth is now 22 kHz!
This model is still in the training phase, so there may be updates in a while.
2023-03-17 02:01:32 +0000 UTC
Announcement - backing vocals extractor
A new model for backing vocals extraction will be ready in the summer.
The new model will be called "b.v.e.". To avoid name confusion, we will soon rename "b.v." model to "kar".
2023-03-09 12:00:02 +0000 UTC
Useful information for subscribers
Some of our users find that they cannot pay their subscription through Boosty. Here are some tips on how to solve this problem.
If the card payment is unsuccessful:
1. Check that you have selected the correct card payment method.
2. If the card is already linked, try entering the details again.
3. Contact the support service of the issuing bank, as the cardholder's
card most likely does not support 3D-... 2023-03-04 13:11:45 +0000 UTC View Post
1. Check that you have selected the correct card payment method.
2. If the card is already linked, try entering the details again.
3. Contact the support service of the issuing bank, as the cardholder's
card most likely does not support 3D-... 2023-03-04 13:11:45 +0000 UTC View Post
What model would you like next?
VOTING IS OVER
1 - piano
2 - strings
3 - synth
4 - denoise
5 - applause
6 - backing vocals extractor
A new MDX (b.v.) v2 model has been added
This is a new version of the model for removing only lead vocals while keeping the backing vocals in the music.
2023-02-03 05:09:53 +0000 UTC