Sine From Above stems + instrumental (best possible AI quality)

[ROSHFORT 98]

I processed with the help of a recently released updated model from MDX, it was trained on files .flac with a focus on the separation of vocals, but it copes with the instrumental too well.
instrumental+EQ
vocals
bass    bass+EQ
drums
other+EQ

uncompressed .wav files and files without equalizer

 

[ARTPOPe 10,016]

That acapella is so clean, holy 

[Noah 26,583]

This is amazing, it's wild how far AI has come! I remember being 11 and trying to get terrible Instrumental Versions of Songs using Audacity. Thank you for sharing! 

[ROSHFORT 98]

17 minutes ago, ARTPOPe said:

That acapella is so clean, holy 

Here is another version of the vocals, but using the ensemble method, that is, I processed the song not through one model, but through 5 and combined the result. It can be even better in some places vocals ensemble method

[MANiCURE 6 10,197]

The instrumental sounds amazing ! could you do the other songs on Chromatica please? 

[ALGAYDO 29,947]

Wow these slap. More Chromatica please 

[HugoJG 1,061]

Could you do the same with ARTPOP (song)? I'm tired of waiting for the stems to leak

[ROSHFORT 98]

Other without EQ

Instrumental without EQ

[Blake 7,385]

This is exactly what Moises.AI does.

[Elliot Parker 353]

GOD IS HERE

[tigerteeth 565]

These are beautiful! Thank you

[ROSHFORT 98]

4 hours ago, Blake said:

This is exactly what Moises.AI does.

https://mvsep.com/quality_checker/leaderboard2.php?sort=other

In order not to write nonsense next time, check the information. Here is a table of results of different models, as you can see Moises is in 20th place, while MDX-23 is in second place after the original stems, in the instrumental category Moises.ai is generally in 137th place, so you not only tried to insult me as the author of the post but also misinform people.

[DISZKO 640]

Wow this is amazing, any chances of  ROM or Alice vocals (with bridge) ? Could you do those if its not a lot of energy to do?

[ROSHFORT 98]

32 minutes ago, DISZKO said:

Wow this is amazing, any chances of  ROM or Alice vocals (with bridge) ? Could you do those if its not a lot of energy to do?

The hardest part was finding this information and understanding, but the process is relatively easy, I will release a guide on audio separation soon. I am new to this and I found spectral analysis and demixing to be very interesting.

[DISZKO 640]

On 5/22/2023 at 9:12 AM, ROSHFORT said:

The hardest part was finding this information and understanding, but the process is relatively easy, I will release a guide on audio separation soon. I am new to this and I found spectral analysis and demixing to be very interesting.

Sounds good, I like your AI stuff. Its cool that you have the patience to see it through haha. I would love to know it as a producer to help with production or making remixes.