Removing music the easy way!

[DonkeyKonga]

Ok, hear me out. I found a great way to easily remove audio tracks and seperate voices! ( or vice versa )
First you use this AI to seperate the voices from the audio:

So I tested all the AI models and I had by far the best voice extraction results with the bottom option called Demucs 4 HT!

Vocal & Instrumental Isolation

MVSEP performs separation of audio into vocal and instrumental parts, extracts text from audio and it is free. Uses Artificial Intelligence.

mvsep.com

Now the voices can sometimes sound not quite right

Then you use this ai program by adobe (free to use) to turn it into studio quality!

Enhance Speech from Adobe | Free AI filter for cleaning up spoken audio

This AI audio filter improves spoken audio to make it sound like it was recorded in a soundproofed studio.

podcast.adobe.com

Results vary, but so far I've had mostly succes to different extends!
For the uninitiated you can also remove all voices from 5.1 audio tracks this way! Simply export each mono track and upload them seperately. Voila, voiceless 5.1 audio WITH effects (using Demucs 4 HT). Name them in the correct order and put the files back in the 5.1 track in your editing software!

 

[DigModiFicaTion]

mvsep is a fantastic tool. Been using it for a while now. I'll have to check out the adobe program. Thanks for the heads up!

 

[addiesin]

Awesome, the first part was just that, part. This is the other half of the puzzle we needed.

 

[Gieferg]

Tried that on one scene that I thought can't be succesfully "un-scored".
Well, I was wrong

 

[DonkeyKonga]

Tried that on one scene that I thought can't be succesfully "un-scored".
Well, I was wrong

Exciting to hear! Ahum, feel free to show the results!

 

[Gieferg]

It will be included in my upcoming Mortal Kombat edit. It is Shang Tsung's line:
"Scorpion and Sub-Zero. Deadliest of enemies. But slaves under my power." There was music playing loud in central channel when he said it.
I am re-foleying this whole scene (and many more) for complete re-score, but needed dialogue lines intact.

 

[DonkeyKonga]

So I tested all the AI models and I had by far the best voice extraction results with the bottom option called Demucs 4 HT!

 

[tremault]

I've been using melody.ml , any idea how they compare?

 

[DonkeyKonga]

I've been using melody.ml , any idea how they compare?

No, but I know this one is really good.

 

[DigModiFicaTion]

Just checked out the adobe podcast tool. I think this works great for when you have background noise left in the mix. I've been able to produce center vocal tracks that are very comparable to the podcast app results. Nice thing is that you don't have a time limit to the conversion like with mvsep.

GitHub - Anjok07/ultimatevocalremovergui: GUI for a Vocal Remover that uses Deep Neural Networks.

GUI for a Vocal Remover that uses Deep Neural Networks. - GitHub - Anjok07/ultimatevocalremovergui: GUI for a Vocal Remover that uses Deep Neural Networks.

github.com

 

[mikeyla85]

Hey all!

I've heard that there are AI tools to remove the backing music from a track. Have you tried them on movies — do they work?

 

[DigModiFicaTion]

Hey all!

I've heard that there are AI tools to remove the backing music from a track. Have you tried them on movies — do they work?

#staff
I merged your thread with this one. The search function can be used to find answers and existing threads on this topic.

 

[Gieferg]

After removing music from whole Mortal Kombat movie I can confirm these are in fact very helpful tools.

It was even possible to succesfully remove music from scene from Mortal Kombat trailer, the one with dialogue: "You can't run from me, Shang! - I dont need to run", this is amazing

That being said, a whole lot of SFX rebuilding was still necessary in several scenes.

 

[The Scribbling Man]

That being said, a whole lot of SFX rebuilding was still necessary in several scenes.

In my experience, there are a lot of good tools for pulling out dialogue from tracks soaked in music. But the SFX almost never fare well enough to use. It's the dialogue that can't be replaced though, so thank goodness it can be isolated!

 

[Gieferg]

Also, rebuilding SFX can be fun

 

[The Scribbling Man]

It can, but too much of it (aka, more or less a whole film) and I find it tedious. I love re-scoring though

 

[addiesin]

I've been using melody.ml , any idea how they compare?

This site appears to be down.

 

[tremault]

This site appears to be down.

Worked for me just now.

 

[DonKamillo]

Ok, hear me out. I found a great way to easily remove audio tracks and seperate voices! ( or vice versa )
First you use this AI to seperate the voices from the audio:

So I tested all the AI models and I had by far the best voice extraction results with the bottom option called Demucs 4 HT!

Vocal & Instrumental Isolation

MVSEP performs separation of audio into vocal and instrumental parts, extracts text from audio and it is free. Uses Artificial Intelligence.

mvsep.com

Now the voices can sometimes sound not quite right

Then you use this ai program by adobe (free to use) to turn it into studio quality!

Enhance Speech from Adobe | Free AI filter for cleaning up spoken audio

This AI audio filter improves spoken audio to make it sound like it was recorded in a soundproofed studio.

podcast.adobe.com

Results vary, but so far I've had mostly succes to different extends!
For the uninitiated you can also remove all voices from 5.1 audio tracks this way! Simply export each mono track and upload them seperately. Voila, voiceless 5.1 audio WITH effects (using Demucs 4 HT). Name them in the correct order and put the files back in the 5.1 track in your editing software!

I've been using the Music Rebalance option in the iZotope RX8 Audio Editor for extracting dialogue in music heavy scenes and was always quite happy with the results. It also lets you separate an audio track into dialogue, music and SFX tracks with very decent results.

But this combination of tools sounds too intriguing not to try out, thanks!

 

[addiesin]

Worked for me just now.

Weird, this is what I'm getting.