Dane 07.11.2019 13:46 |
A couple of days ago Deezer released their open source code Spleeter which uses deep learning algorythms and artificial intelligence to seperate Vocals from Backing Tracks of any song. It separates a maximum of 5 stems; vocals, drums, bass, and 2 others (haven't tried yet) Ofcourse, this is only in its early stage but my first result sounds promising. |
innuendo73 07.11.2019 14:02 |
Imagine being able to remove “property of Queen Productions” |
matt z 07.11.2019 14:49 |
That was interesting but flawed as it never removed enough of the original source. It DID make me appreciate the soundscapes they'd created even more - simply hearing that isolated guitar trill... that was beyond interesting. Stems would be amazing. |
AB-88 07.11.2019 15:05 |
Would you get a better result if you used the 5.1 stems from the German Greatest Flix DVD? |
dysan 07.11.2019 15:22 |
Great, more vocal snippets for Freddie Goes To The Shops memes. |
gambri 07.11.2019 17:37 |
Amazing!!! Thanks |
mooghead 07.11.2019 22:03 |
Nah.. a ghost hovering above your ears... nothing revealed. A bit unsettling actually... |
stevelondon20 07.11.2019 22:58 |
Phenomenal sound! |
brians wig 07.11.2019 23:35 |
Can it remove the vocals and leave the rest? |
flash00. 08.11.2019 04:02 |
That's amazing! some voice voice Freddie had.. I thought you did a good job cheers. |
musicland munich 08.11.2019 04:24 |
We will see a CGI Freddie in the future moving/ acting to concerts that never shot on film. Scary and fascinating at the same time. |
SweetCarolina 08.11.2019 08:30 |
How do you do it? |
stevelondon20 08.11.2019 09:08 |
You chose the best Queen song too! Lol. Well in my opinion anyway! |
mooghead 08.11.2019 11:18 |
brians wig wrote: Can it remove the vocals and leave the rest?That would be a lot easier technically. |
Dane 08.11.2019 13:35 |
This process outputs (in my case) two files in a new folder. 1. vocals.wav - not the main vocals but all vocals in a song 2. accompaniment.wav - all the rest If you extract the maximum of separated files you get the following 1. vocals.wav 2. bass.wav 3. drums.wav 4. piano.wav 5. accompaniment.wav (which is all that's left) As soon as I have it up and running again (there's a codec issue as of yesterday) I will try extracting all five stems from Innuendo. |
Dane 08.11.2019 13:42 |
But for now... |
Ale Solan 08.11.2019 14:01 |
AB-88 wrote: Would you get a better result if you used the 5.1 stems from the German Greatest Flix DVD?huh? |
Dane 08.11.2019 14:53 |
OK, fixed the issues and created a mash-up of all stems that were generated. |
GAP 08.11.2019 19:19 |
Wow! Thanks for sharing these gems and also for dedicating lots of hours in this project. It could be possible to try with these master pieces: TMOTBQ, TFFMS & SAVE ME, please. |
MrFunster 08.11.2019 19:36 |
All Gods people. For me also a true masterpiece. It seems is often forgotten. Very powerful. And i love to get a vocals only for that one. |
dysan 08.11.2019 20:37 |
Dane wrote: OK, fixed the issues and created a mash-up of all stems that were generated.It's a bit messy. (edited out harshness of initial reply) |
SweetCarolina 08.11.2019 20:44 |
I want to do the same! How can we do it? I have Deezer but where is that program? |
TheAdmiester 09.11.2019 00:54 |
This was a pain to set up but definitely worth it. It's not perfect (instrument bleed through, watery sounds where frequencies were removed), but it's insanely good, better than anything else I've ever used and likely the best we'll ever get without real stems. Here's Days of Our Lives, vocals only since we've already heard an official instrumental before. https://sndup.net/8hpb/vocals.mp3 |
Wilki Amieva 09.11.2019 05:40 |
What if we train this using pre-mixes from Rock Band/Guitar Hero/QEx/competitions/etc stems, multi-tracks, instrumentals and acapellas? |
BETA215 09.11.2019 06:39 |
It's doable, but we need more coding skills. Setting up the program might be a bit hard, but doing THAT is another level of skills. The thing isn't doing amazingly complex things, but more like knowing different kind of langs/structures. ">link BTW, still love your One Year Of Cris cover. |
dysan 09.11.2019 07:07 |
There's a bit of software (I can't remember for the life of me what it's called) and we were mucking around with the free trial version of it in the studio and it actually works for this stuff. I was hopeless at it but my pal was isolating vocals cleanly from random scratchy mono stuff he found on youtube to test it. I was genuinely staggered. I'll ask him what it's called. RX or something. |
kosimodo 09.11.2019 07:19 |
Never boring indeed. |
Sebastian 09.11.2019 08:54 |
Really interesting and fascinating. Thank you. Sure, it's not perfect, but it's way better than nothing at all. |
Wilki Amieva 09.11.2019 10:26 |
...That would be Izotope RX, a great piece of software for audio repair. It includes something called Music Rebalance which is tweakable, but I think this approach is better, especially if you can re-train it for specific content. The good news is that I use Python more and more each day (including stuff for audio), and I also do work with AI for satellite image processing. I also have my own recording studios. I might be able to handle this! The bad news is that creating the training/test database would take a LOT of time. |
Star* 09.11.2019 11:52 |
Enjoyed listening to those pieces of Queen great! Love to have a Cd of vocal only Queen tracks. "All gods people" will be very special to listen to though ! |
Wilki Amieva 09.11.2019 13:06 |
I was curious whether this can be applied to live recordings... So I have run the process through Need Your Loving Tonight from the first Buenos Aires concert, I did some post-process to recover some of the lost stuff and then I did a quick remix of the track. This is just a raw mix with some minimum FX: link And karaoke (not perfect): link The results look promising! P.S: I should have corrected that wrong note from the bass. |
Chrisallstar 09.11.2019 13:53 |
That’s sounds great, could you do a full instrumental of “Innuendo” from start to finish? |
gambri 09.11.2019 17:41 |
Wiki, it sounds amazing !!!!! Excellent work!!!!! I'm not an expert in mixing, maybe the voice is a bit low? I wonder if with this soft you could raise your voice at the concert of the Golders Green Hippodrome 73? |
gambri 09.11.2019 17:46 |
Sorry for my Google english. Obviously I was referring to the Freddie's voice. |
dastard 09.11.2019 19:17 |
Sounds great. I would love to get my hands on the stems from The White Queen. I have tried to use this program but not getting bit to run. |
Dane 10.11.2019 14:09 |
@Chrisallstar here you go. |
advanced159 10.11.2019 18:35 |
The Millionaire Waltz Band link And Vocals link |
mooghead 10.11.2019 19:07 |
Take away the 's' from https and the links will become clickable and at least get 10 trillion more views.... For the trillionth time. |
dysan 10.11.2019 19:58 |
This is essentially the same as the old bootleg called the Golden Demos where a bootlegger had mucked around with the left and right channels to remove singing. Essentially I don't think any good will come of this. |
dysan 10.11.2019 20:01 |
(don't get me wrong though - it IS interesting when done well) |
Wilki Amieva 10.11.2019 22:29 |
dysan wrote: This is essentially the same as the old bootleg called the Golden Demos where a bootlegger had mucked around with the left and right channels to remove singing. Essentially I don't think any good will come of this.You couldn't be more wrong! This is a totally different thing, with endless possibilities. |
Wilki Amieva 11.11.2019 00:18 |
By the way, I updated the Buenos Aires 1981 files. |
dysan 11.11.2019 07:34 |
Wilki Amieva wrote:How so? Fans taking a track and playing with it to make something else. Fine. I get that the technology is fascinating, but to what end? Someone will monetise this and dilute the 'unreleased' pool even more than it already is.dysan wrote: This is essentially the same as the old bootleg called the Golden Demos where a bootlegger had mucked around with the left and right channels to remove singing. Essentially I don't think any good will come of this.You couldn't be more wrong! This is a totally different thing, with endless possibilities. |
BETA215 11.11.2019 07:55 |
This is artificial intelligence, which learns from material it's given. It can learn from audio and recognize groups of frequencies related to vocals, bass, drums, guitar, etc. Of course, the initial result of this is something, heard alone, uninspiring. If it learns to divide correctly vocals from all the instruments, even being perfect, it won't sound absolutely perfect. But it could be usable for improving already existing material like: badly mixed songs, good Queen concerts recordings with bad or not good enough mixes (Golders Green, Argentina), delete noise and unwanted sounds without damaging actual useful data, improve in different ways already existing material, etc. |
BETA215 11.11.2019 08:02 |
I don't really care whoever monetise this. Of course it's terrible, but stopping advances in terms of its consequences in the black market, it's something not THAT important to me. I mean, are Brian May, Roger Taylor or John Deacon poor? Will this affect their income? Do they care about what we want to hear? Obviously no. Then, if the band won't give us what we want, it's not bad to reach it in our own ways. I mean, we're not doing piracy... And if it enables it, it's part of 2019's life. We (neither QPL) can't control that kind of business. Else, we should be questioning ourselves, do my actions in my every day life contribute to inequality? Should I stop buying technological devices? Stop buying from the supermarket? Stop buying CDs? Everything can enable unfairness, even doing nothing regarding anything. The thing is helping when you can, and living when you can. |
Dane 11.11.2019 12:00 |
OK, so at request I have attached White Queen in 5 separated tracks. I haven't even listened to this myself as I am currently on holiday but I trust it's interesting. As for the 'use' of this technology. I imagine it is very usefull to clean up and remix/remaster recordings from the early sixties and before. This is where only 4 track (or less) multitracks exist and in order to separate any individual instrument or vocals, say, the first Beatles albums this will work a charm when educated correctly. As for monitizing this, there will always be assholes fucking things up for the rest of us. Just make sure to download the (currently) open source version of this software as long as it's available. |
dastard 11.11.2019 12:07 |
Dane! Thank you so much, it is so great to hear this! |
dysan 11.11.2019 13:45 |
@beta you first response was good, the second went off piste. |
Chrisallstar 11.11.2019 14:38 |
Thanks so much Dane for the Innuendo instrumental |
The Real Wizard 11.11.2019 15:24 |
Wilki Amieva wrote: I was curious whether this can be applied to live recordings... So I have run the process through Need Your Loving Tonight from the first Buenos Aires concert, I did some post-process to recover some of the lost stuff and then I did a quick remix of the track. This is just a raw mix with some minimum FX: link And karaoke (not perfect): link The results look promising! P.S: I should have corrected that wrong note from the bass.Indeed, it ain't perfect - but it's impressive that even this is possible. We're in the space age now ! |
The Real Wizard 11.11.2019 15:31 |
Dane wrote: OK, so at request I have attached White Queen in 5 separated tracks. I haven't even listened to this myself as I am currently on holiday but I trust it's interesting.It's mighty impressive. I stand in awe at both the results and the potential for what's next. Edison would be pleased. |
dysan 11.11.2019 17:58 |
Parts of the 'other' track (especially the first 100 seconds) are genuinely haunting. The melody and the digital artefacts merge somewhere that transcends the piece. |
Holly2003 11.11.2019 18:37 |
The Real Wizard wrote: Edison would be pleased.I think he would be too busy electrocuting dogs, cats, horses and elephants to notice. #Justice4Topsy |
emrabt 11.11.2019 19:51 |
I miss-read the thread title and thought this was going to be a video where Adam Lambert talks us through recording vocals on multitrack. |
dysan 11.11.2019 19:53 |
haha |
thomasquinn 32989 12.11.2019 09:45 |
The Real Wizard wrote:Edison would be looking for a way to steal it.Dane wrote: OK, so at request I have attached White Queen in 5 separated tracks. I haven't even listened to this myself as I am currently on holiday but I trust it's interesting.It's mighty impressive. I stand in awe at both the results and the potential for what's next. Edison would be pleased. But I totally agree that the potential for this kind of software is incredible! |