Isolating audio from a tape with accompaniment

Hello everyone. I have googled this question but nothing quite matches what I'm trying to achieve and any advice would be really appreciated.

I have a cassette tape of my late Grandmother singing over the top of an LP which was then recorded (probably in her lounge). I started getting into amateur music production myself about 18 months ago and have had to teach myself (with the help of others on youtube etc) about DAWs and song production. I have a Maschine MK3 and an S series MK2 keyboard. I also have Cubase LE/AI Elements 8 and Ableton Live lite.

Thus far I have managed to play the songs she sang into the MK3 as samples and chop up said samples into workable chunks. Unfortunately I can't use them like this as you can still hear the orchestral accompaniment and (probably) Bert Humpadink or whatever his name was singing with her! Ideally I'd like to provide the accompaniment myself with hers being the only voice heard.

As I've said, I've googled this and the only examples I've found on youtube show audio extraction from digitally produced tracks which won't work with my samples. Is there a way to do this or should I just admit defeat and send the tape off somewhere for someone professional to do it for me? (No idea where to start on that either).

Calling all super brainy boffins please!

 

it will be allmost impossible for you to isolate her voice if there's a lot going on in that recording. you'll allways have to sacrifice some/a lot of her "sonic quality"...

meaning: i am sorry to say it --> you'll never get really good results, because everything you / others do to your audiofile will involve a lot of filtering / eq'ing and so on...and will problably just sound bad/thin/not-worthy in the end...

i feel your pain. i've lost lots of recordings which i miss now. but that's life i guess...

good luck to you anyway.

 

I would add. It also depends what is the volume of "background" (LP, 2nd voice, ... ) comparing to main voice. And the quality of the recording. If it is on cassette, I guess the quality of recording is not very high. If not rather poor. In that case the filtering, equing and so on may not harm the result that much.
Generaly said, the task is not easy, unless the main voice is much stronger than the rest. At least for most of time.
I guess it is more the task for someone working for police or researchers at university (e.g. research on AI understanding spoken language in real environment must have means to filter off unwanted background sounds.), than musicians or sound designers.
But in any case, you can hardly count on that the result will be "main voice in great audio quality". Unless the recording was taken with good microphone on hi grade cassette recorder in sonically treated room. Which is not, most probably, the case.
Maybe, if you upload few chunks of the recording, someone could guess how hard (and expensive) it would be to separate the main voice.

 

Thank you both for your replies. My Nan had one set of lungs on her and I'd say the loudest thing on the tape! She had a self-taught operatic voice. I will see if I can upload a bit of her voice for you to listen to...

OK, well I've saved a WAV file to my desktop of a snippet of her singing but I can't upload the file here, it says it doesn't have an allowed extension so that's a pain.

 

You can pack it in a zip or rar container.

 

yes try to upload it as a zip/rar or put in on dropbox/icloud/google drive or whatever.

i add a little bit of "hope": i re-read your post and i only now realised that she was singing over an LP that was probably runinng in the background.
your biggest chances one can help you will be if you can provide a recording of that same LP she was singing to. that way we could try the "phase inversion" method. meaning we can align "her recording" with the pure LP-song in an editor and swap phase on one of the files....as long as it's the exact same LP-song chances are that we can "null/cancell" everything but the additional recording of her voice/room reflexions...

but don't hold your breath "too deep" because cancellation methods usually thins out a sound A LOT...

but maybe this way you at least can have her voice "allone" (kind of) and do your own singing along her..i'd wish that this somehow works out.

let's see how your file sounds and if you can find the pure recording of the song on the LP.

good luck.

 

Right, just learned how to zip a file (yes, I know, lots to learn!) Here goes....

Will have to ask family for original LP, that could be tricky but one step at a time...

Thank you all for your interest and help in this by the way. I don't sing but am a reasonably accomplished pianist so it would be lovely to build a track with Nanny as the star.

 

...if i find time tonight i try something with it (no promise). but: do you at least know the title/band of the music "in the background"?...
but as i said earlier: i doubt any of the results we may come up with will be good enough to have a "pop style vocal" that could classify your Nanny as a star....let's see.

(btw: did you know you posted this in the not really well frequented german section of that forum?...
i try to bring some traffic in here by posting your thread on the english part and put a link to this very thread here).

 

What a twit I am! How on earth did I put this on the German version of the forum. Sorry! I admit I had a problem creating a post to begin with - I think I started by doing it as a conversation and then realising this was basically a private message. Oh dear...

Thank you for helping me out, as you can see I really need it!

 

I guess, the LP is not needed physically. The name of song and LP might do and maybe release date. The cancelation of the background using song from LP is very good idea. But it will not be that straitforward. Recording might be at a bit different speed, the speed might float in time, and the song on cassette recording will be as if filtered by frequency filters, .... To allign cassete song and LP song in time I would recomend using correlation function that would be computed dynamicaly for each (each tenth, ...) sample point. That would show, if speed is lower or higher and if and how it floats in time.
But definitely, using the song from LP is the step in right direction, I guess.

Thanks for recording. The voice is pretty strong, comparing to LP. I guess, it should be possible to separate the voice more or less.

 

If your daw or wave editor has a karaoke function use it and try the "phase inversion" method on the karaoke result on track two. On track one is the nanny's record.

 

good morning pandora. i've had a listen yesterday evening.

first: your Nanny's got a beautiful voice, i understand now why you want to try to "rescue it"

good thing first: the background music luckily is really quiet in the examples you posted. all our talking above was under the assumption that the voice is "burried" completely in the background track - like in a regular recording.

my suggestion: ofcourse this voice-samples can't stand on their own in a musical piece. but with together with other elements i think you coud be fine with just a bit of equalisation of the voice-sample.

i would not recommend you to send this files everywere to get it "cleaned" for money. as long as there is no musical context to put this samples in, its useless to start EQing it. in a full mix the "background noises" would get covered by your pianoplaying, strings or whatever...

and the mentioned "phase inversion" methods wont bring really better results then just filter out the unwanted frequencies with an EQ. biggest "problem" will be that your Nanny sings "sustained" in the examples so we can't find a spot in the sample where there's only "backgroundmusic" which we could try to use in a "noise removal tool". if you've got a recording / other examples where we only can hear the quiet background-music-part we could try something...

but for now that's all i can say: try to compose something along her singing and fix the "background thing" in the mix.

it will never sound as good as your Nanny still alive singing freshly in a microfon. but this recording is beautiful enough and deserves you do something musical with it - even when that means without perfection