Traktor Stems: Is sum of all four tracks identical to stereo version?

I've been doing some comparisons between a 3rd party stem creation tool and Traktor's new built-in stem tool. Something that caught my attention with this other tool is that they claim that when their drums+bass+other+vocal tracks are played together, the resulting sum is bit-for-bit exactly the same as the original stereo track. (They show a test by playing the stereo vs. the summed stem out-of-phase and having it 100% cancel.)

To my ears, it sounds like the Traktor-made stem files have the same quality — that is, if you play all four stems together it is indistinguishable from the original stereo source.

Is this the case? Has anyone heard a claim (or done their own tests) to prove that you lose no quality when you are playing back a stem-analyzed file (when all four stems are active)?

Find more posts tagged with

Comments

canton

Just a follow-up to my own post. I pulled a Traktor-made .stem.m4a file into my DAW and saw that it has 5 tracks: drums+bass+other+vocal and then the original mix. Summing the four stems and playing them against the original mix definitely had loss (lots of stuff left over when inverting the phase vs. the original).

Does anyone know in what situations Traktor might play that original stem, the 5th track in the .m4a? Like, if you enable all 4 stems and don't adjust the volume or FX on any of them?

ErikMinekus

The 5th (master) stem is used by audio players that do not support the stems format, I don’t think Traktor ever plays it.

The stem creator tool does have some EQ options that Traktor uses to make the stems sound more like the original, but I doubt it can be bit-for-bit.

canton

https://community.native-instruments.com/discussion/comment/193781#Comment_193781

Here's the thing though: I did a test to compare the original stereo file played through Traktor vs playing the 4 stems combined but unmodified. When I phase cancelled these, there was just the tiniest difference between the two — a very very quiet version of the track, but no inconsistencies in the frequencies.

Look at this image: zooming in on the waveform, they look identical, down to the sample. It seems like only a difference of amplitude, maybe something to do with my having forgotten to turn off the limiter for this test.

However, when I manually combine the four stems in my DAW, the resulting sound is really, really different from the stereo file. Much ******.

So my conclusion is that the reason Traktor stems still sound so good when you have all four stems playing is because in that case it actually switches over to that 5th master stem. (This would also be smart from a performance perspective. Why pitch shift and spin up 4 stems when you could just play the stereo instead?)

Quick links

Unanswered questions