MODEL for - De-Echo / De-Reverb VOCALS ?

[AlonDan]

Hi All, Is there a recommended MODEL to reduce Echo / Reverb for VOCALS ?

Thanks ahead!

[thecatontheceiling]

no, this is a vocal remover, not a de-reverber

[AlonDan]

Thanks for the reply Since MDX-NET model also include a DeNoiser (DeNoise option) Maybe other models include DeEcho / DeReverb as well.

[thecatontheceiling]

I literally just told you that there isn't a de-reverber. please refer to other tools on the internet for that. this is a vocal remover, not a reverb remover.

[AlonDan]

you literally didn't understood me, which is fine. I wasn't talking about the TOOL, but the models it uses to remove / separate vocal.

Just information for anyone who's interested in the same thing: Since vocal Remover in some models as I mentioned above also include "what to remove or how" for example: the EXTRA OPTION in MDX-NET model to remove noise a.k.a. DeNoiser, at the same: DeEcho / DeReverb it's a training architecture that is possible to include in a MODEL.

Based on a long discussion on discord with some talented people who train models via Demucs and MDX-NET I understood that not only it's possible to accomplish, but some of the guys is training his first MDX-NET model to support DeNoiser + DeEcho which will defiantly take longer time to render (for users) based on what he explained. So I received my answer but I understand there are no public models with that option yet, but I'm glad that it's on the working I just hope it will be public so we can enjoy this great extra option.

[thecatontheceiling]

??? I told you the models didn't support de-echo and the tool doesn't either. compatibility with such models is not coded in UVR itself yet.

[nukmuk]

Reverb467 and delay_v2_nf2048_hl512 from Discord work pretty well for me.

[AlonDan]

Reverb467 and delay_v2_nf2048_hl512 from Discord work pretty well for me.

Thanks for sharing! More advanced options models are coming with these which is very useful.

[dts350z]

Reverb467 and delay_v2_nf2048_hl512 from Discord work pretty well for me.

Can you say more about the usage? How to integrate with UVR or how you otherwise used the models. Are there instructions posted somewhere?

[KatieBelli]

Reverb467 and delay_v2_nf2048_hl512 from Discord work pretty well for me.

The delay model: How do I import it successfully into UVR 5 GUI? It's a pth file and I had no issues importing the reverb model but I'm stuck with the delay model. Looking forward to your help. :)

Edit: Tried it today and the delay model seems to be not working well for my intended purpose: removing echo/delay.

[Anjok07]

A really good Reverb model created by a UVR user has been added to the Download Center in the GUI.

Settings Menu -> Download Center -> MDX-Net -> Reverb HQ By FoxJoy

[KatieBelli]

A really good Reverb model created by a UVR user has been added to the Download Center in the GUI.

Settings Menu -> Download Center -> MDX-Net -> Reverb HQ By FoxJoy

A wonderful model, thank you! :)

Is there a website or Telegram/Discord group which shares similiar models? I'm looking for a model that can remove echo/delay. The one that's been posted here doesn't work well on my audio files.

[a-ggghost]

A really good Reverb model created by a UVR user has been added to the Download Center in the GUI.

Settings Menu -> Download Center -> MDX-Net -> Reverb HQ By FoxJoy

The new models are fantastic! Is there a way we can swap the primary & secondary outputs of Reverb HQ?

The reason I ask is because the dry vocal from Reverb HQ is the secondary "No Other", but dry vocals from UVR-De-Echo-Aggressive and UVR-De-Echo-Normal are the primary "Other". It prevents properly ensembling the De-Echo models with Reverb HQ except manually.

[Anjok07]

This has been fixed! They should be swapped now.

[Dyslexicon]

Dummy question - I'd like the entire mixdown cleaned of Reverb but dont need/want any stems separated; which boxes should I have checked using ReverbHQ to just process the file and spit out a de-Reverbed copy? I just ran it and ended up with 2 files: a de-reverbed mixdown of everything but guitars (other), and a guitar stem still heavily laden with reverb.

[RobertAgee]

You'd probably want to separate the stems individually, then dereverb them, then mix them back together.

[980202006]

Hi, I want to train a new dereverb and deecho model use my dataset. Is there any details?

[thecatontheceiling]

??? I told you the models didn't support de-echo and the tool doesn't either. compatibility with such models is not coded in UVR itself yet.

Uh, yes it is, been a thing for a while...

for your information, that comment was made 7 months ago now it is in uvr and it's pretty cool

[801artistry]

no, this is a vocal remover, not a de-reverber

there has been and is, what the hell? lol de-echo, reverb and mdx reverb hq

[thecatontheceiling]

that reply was made a year ago. now there exists very good model for all three purposes.

[BlannyBlannyBlanc]

that reply was made a year ago. now there exists very good model for all three purposes.

Do enlighten me, because I'm curious now where such a model exists.

[LittleCyberCortex]

Do enlighten me, because I'm curious now where such a model exists.

If you don't see any type of those models anywhere in the three categories of your installation, just browse the downloads The names of the listed special purpose models are usually self explanatory

[thecatontheceiling]

that reply was made a year ago. now there exists very good model for all three purposes.

Do enlighten me, because I'm curious now where such a model exists.

sorry, i dont use UVR anymore :P