What are your methods for transcribing individual instruments from mixed audio?

Chords, scales, harmony, melody, etc.
User avatar
13206 posts since 8 Mar, 2005 from Utrecht, Holland

Post Tue Oct 19, 2021 10:24 pm

The music style influences the method to use.

What I do is write down the chord progression & structure. I find melodies easier to "just remember" and play than to jot down in staff. Does not help in the long run, I do forget stuff.

Lead sheets... You can buy books with transcribed pop songs for guitar or piano. Browsed through some, always found errors or blatant simplifications that give wrong results. Some say this is done to avoid copyright, but I don't buy it (sic)

How to check you got it right? I used to input the chords into BandInaBox and let it play it. Then to find out the closest style insists on playing all chords with the seventh which it should not do in that case. Oh well...

I use my ears, nothing else. Its a gift. If stuff is masked, its not important for transcribing.

What's more important is the ability to write down what you hear in your head, without having a recorded example. I imagine things my fingers cannot play. Should practice more often.

Here's a story I recently found Bruce Swedien told about Quincy Jones
SoundOnSound.com wrote: Quincy and I first worked together with Michael Jackson on the movie The Wiz. We were living together at a hotel in Manhattan, and we would go to Studio A at A&R Studios. We had a big session at noon on Monday to record some of the music with a big 70‑ or 80‑piece orchestra, and we had to leave for the studio at 10am. The night before, Quincy and I had guests at our hotel for dinner, and Quincy still hadn't even started on the orchestration for the opening titles. I was getting a little nervous, but he said not to worry about it. At about four that morning, I woke up and noticed under my door that all the lights in the apartment were blazing. There's Quincy at the dining‑room table with a billion sheets of manuscript paper, and he was writing orchestrations. I said 'Quincy, we've got to leave soon!', but he just said 'Don't worry about it' so I went back to bed.
At about nine o'clock I got up again, and Quincy said to me 'I'm all set'. There wasn't even a piano or a guitar in the apartment; just Quincy and his manuscript paper! Off we go to the studio, and Quincy hands over his score to the copyists. He didn't even want to conduct — he'd hired a conductor because he wanted to be in the control room with me. The conductor gave the down beat, the orchestra played the entire overture, and there was not a single note out of place. It still gives me the chills to think about it!
We are the KVR collective. Resistance is futile. You will be assimilated. Image
My MusicCalc is served over https!!

232 posts since 4 Aug, 2020 from Montreal, Canada

Post Wed Oct 20, 2021 5:26 am

BertKoor wrote:
Tue Oct 19, 2021 10:24 pm
Here's a story I recently found Bruce Swedien told about Quincy Jones
I figured that all the supermen or superwomen don't have to sleep and they work at full speed with laser focus anytime!

I enjoyed reading that book. Inspirational. Legendary.


Topic Starter

92 posts since 26 Mar, 2017

Post Thu Oct 28, 2021 10:07 am

donkey tugger wrote:
Tue Oct 19, 2021 12:51 pm
N__K wrote:
Tue Oct 19, 2021 12:04 pm

I'm vaguely aware that nowadays there are AI-based demixing and transcription tools. Do tell if you know of such tools and how to use them :)
I know knack all about music theory, but on the de-mixing side this is free and seems to do a job (Windows only), if a bit rough and warbly. Only messed about with it out of curiosity so I'm sure there are tricks you can use to get better results etc.

Thanks! I've seen Spleeter mentioned before, but hadn't tried it until now, due to fearing potential technical hassles.

I eventually got SpleeterGUI V2.7 to work on my PCs, and after testing it for a while, the demixing results for music with vocals were better than I expected. In some cases it was rather amazing, despite audible artifacts.

I also tested it on instrumental electronic music - including some for which I had DAW projects - and was positively surprised by how well it separated pitched instruments from drums (separation into 4 parts + recombination of "bass" + "other").

In both cases, while the results weren't always good enough for remixes etc., they were definitely usable for transcribing. I'd go as far as to say that in its current form, Spleeter is probably more useful for demixing for transcription, rather than demixing for remixing.

Interestingly, in separating bass from other pitched instruments, Spleeter appeared to have some challenges in differentiating upper bass partials from treble fundamentals and first few partials, similarly to what I've written about in previous messages in this thread.

But overall, it has now become another tool in my transcription workflows :)


In the future, I hope tools like that could be trainable per-genre, for even more accurate separation of parts; which, in turn, could be audio-to-MIDI'd via apps like Melodyne DNA or AIs trained per-instrument. That'd be of major help in preserving notation/MIDI of music for which no original scores, DAW projects etc. are available.

163 posts since 18 Jan, 2014

Post Thu Oct 28, 2021 11:09 am

Return to “Music Theory”