What's new

Which is the better: MIDI mixing or Audio mixing?

I assume that many users don't bother to create audio files first, but instead create a mix directly from your Virtual Instruments.

I am promoting the more tedious way here.
Why?As a former "WIN-Logic 5.5 user" I used to mix all my projects without audio files to save hard disk space, which was understandable around 2000. But today all those projects are useless. Logic gone, music gone. If I had audio files I could remix them, reapply them etc.

All those who don't make audio files - of every single instrument...without effects - are badly off, if later on some license is missing, a later computer is set up differently, software is not available at a later time etc. Think about this topic in "10-15 year steps". How much can usually change there!

Therefore:
1) So simply read out an audio track from each instrument and save it without any effect. Then create your mix from it.
2) Since I got into the habit of working according to point 1) anyway, I noticed a second advantage: When mixing with audio files, you have to finish the "making music" part once (midi part). If you create the mix from the midi file, then you mix, compose, master, compose, mix... Because you always and everywhere have the possibility to intervene, the project never really ends. This is not very professional, because the professional doesn't have the time to work like that...

Edited: ...Unless you are a professional who can resist the endless changing and also comes to an end in Midi. ;)
(See JohnG's answer below)


Beat
 
Last edited:
Those are good points.

Audio mixing offers a number of advantages:

1. Preservation -- As @Beat Kaufmann wrote above, audio is more likely to preserve longevity of mixes than midi-only;

2. Focus -- many people prefer to mix in audio in order to 'switch hats' more emphatically from composing to mixing brain;

3. Facility -- some kinds of mixing tricks (reprinting, offsetting and distorting a track, then turning it way down as a kind of double-tracking, etc.) are possible to do with midi, I guess, but 10x harder and kind of pointless; and

4. Live players -- even if it's just you chanting (pretend Buddhists?) or clacking something (tin cans, banging a box with a stick) adding anything live often dramatically improves the sound of a cue. If you can play a flute / recorder / trumpet / cello, so much the better. And of course you have to go to audio with that.

Why stay with midi?

5. Longevity not that helpful
-- I almost never revisit cues 10 years later. Why would I? As a media composer, I almost never own the music or the master recordings anyway. Consequently, the only editing I'm likely to do would be for a demo. @charlieclouser would likely point out the benefits of being able to use stems on a sequel.

6. Endless revisions -- media composing, especially today, must be conducted (get it?) in an environment where producers and directors have come to expect almost never-ending changes as the picture / game gets re-cut over and over again. In those conditions, audio can still be great, provided you can edit the stems yourself or you have a music editor. So for deadline-heavy writing, I stay in midi as long as possible.

At the end of the process, even if it's just going to be electronic sources and even if I have no expectation of using the material down the road, I still print everything, for @Beat Kaufmann 's reasons.
 
Last edited:
Excellent advice from Beat, as always. One should definitely render all their tracks for posterity.

But as for the actual mixing process itself, I keep a lot of my tracks unrendered because I use CC controller data to mix.

As a mixing engineer, which used to be my day job, I can't count how many times I have wished I could go back and tweak a performance during the mix.

Just last night I had a piano part that wasn't loud enough in the chorus. Instead of automating the fader, or compressing it, I went in and scaled up the velocity for that section of the song. Like being able to go back in time and tell the piano player to bang a little harder during the chorus.

Not only did this increase the volume, but it gave the piano part more energy as those higher velocity layers opened up.

Other stuff, however, I print pretty much immediately during mixdown. Things like triggered loops that misbehave when you don't start the song at a certain point. Very annoying. So I'll print that to audio just for convenience.

But for most tracks I like to keep all of my options open and only render them out when I'm done.
 
+1 to all the comments so far... I think even if you never render your audio tracks for whatever the reason, mix the audio through the mixer faders, not CC7. Partly for some of the reasons already mentioned, but a few more things to consider:

  1. Mixer faders have much more resolution then 0-127 as in the case of CC7

  2. CC7 is basically turning the volume of your instrument up and down. That effectively reduces the dynamic range of the instrument! You really want your instrument pretty much always set to CC7 volume all the way up...so that the full dynamic range of your instrument is available as the instrument intends it.

  3. Use CC11, Velocity crossfading, etc..to achieve instrument performance dynamics depending on how the instrument itself is programmed, but avoid using CC7 because of point#2 above.

  4. If you need to adjust the actual mix itself in some way..then use the mixer faders....higher resolution and it retains dynamic range of the instrument itself, the level is being adjusted by the DAW mixer.
Regardless of whether you're playing back instrument tracks, or playing back a recorded audio track, the audio ends up in the DAW's mixer as audio. Mix it all there.
 
I render all midi tracks to audio, but keep all the original midi tracks, without any vst instrument loaded.. in the event of having to change a passage, or find a better sounding instrument, I will activate the midi for that part and "render in place". To me, working with audio is much better, reliable and gives better results. Until midi isn't rendered there's always the chance of something going wrong, glitches or any of those incoherent things that computers like to do every now and then. With audio I can really take care of the mix, fx, xfades, dynamics etc. From that moment on it feels like painting to me, choosing colors, refining things... not that I'm paricularly good at mixing, but that's how I like to do it.
 
I always prefer to mix with as many elements as possible still running "live" as MIDI triggering whatever the instrument is. Sometimes I will Freeze some tracks to prevent any random elements from being different on each playback pass, or in the case of super-long samples which I'd prefer not to be forced to play from their beginning in order to hear them (long risers, etc.).

I always prefer to be able to dive into the actual data that makes the track and make changes instead of being stuck with whatever was "good enough" at the moment I made the Bounce / Freeze. When printing final mixes of cues I frequently stop the pass halfway through because I heard something that needs fixing, and dive in to make some tiny changes to note velocities / lengths / etc.

This way of working isn't weird for me at all, since for many years I made records with no multitrack machine - all live MIDI sequences being printed to DAT tapes or whatever. We even scored The Equalizer series in the eighties this way. No audio recording of any sort at that time - just run it live and print the mix. I did it this way when making bigger records as much as possible too, even when we had bigger multitrack setups - although in those days it was a lot more complex and tricky than it is today, since we had no software synths or virtual instruments with total recall, and it would have been 16 or so audio tracks playing from the DAW plus a few dozen hardware samplers and synths. So you needed to make sure every piece of hardware was on the correct preset and doing what it was supposed to, and of course you needed a big console and a lot of cabling! Using this process was less likely when working with an outside mixer, when you'd get everything up and running and then leave them to it, hoping that nobody bumped a knob or whatever without anybody noticing. But even when we printed to tape, 46 tracks was often not enough. On some of the tracks on the NIN album The Fragile we had 46 tracks playing from two Studer 24-track machines, plus a bunch of audio tracks playing from Logic on the ProTools/24 rig, plus some synths and samplers running live off MIDI coming from Logic, simply because there weren't enough audio tracks in the room to handle it all. Then, after finishing, we'd take the reel off the slave machine and put up a third reel, and print all those excess elements to that slave reel. But we didn't have three 24-track machines to play it all at once, so.... gotta do what you gotta do.

And of course, as @JohnG mentioned, leaving as much as possible in its raw MIDI form means you can load up a cue at some later point, do a Save As, and make a new piece of music out of the bones of the earlier piece. When doing a tv series or movie franchise that runs for years and years, this has been invaluable for me. A whole lot of cues in my world have fewer than eight tracks of audio but dozens of MIDI tracks, and since most of those MIDI tracks are triggering EXS sampler instances in Logic, I can get it all back years later and everything sounds exactly as it should. Last month I loaded up a cue from 16 years ago to make yet another version of the end theme for a SAW movie, and I did need to make some tempo changes to one section. Fortunately there were less than a dozen audio tracks (some filtered percussion loops and live strings) and the rest was MIDI > EXS. I used Ableton Live to re-render the audio parts at the new tempo, and I was glad I didn't have to do it to more than a few tracks, that's for sure. All the EXS tracks were good to go, and I could dive in and move notes around, record new parts, etc. as if I'd stepped into a time machine to 2004. I actually copied the MIDI tracks and their Channel Strip settings into my current template so I could add elements from the sounds I had used on the rest of the film and output everything through the current stem matrix and effects. This took an hour or so but was worth it.

For me, I just couldn't do it any other way.
 
I render all midi tracks to audio, but keep all the original midi tracks, without any vst instrument loaded.. in the event of having to change a passage, or find a better sounding instrument, I will activate the midi for that part and "render in place". To me, working with audio is much better, reliable and gives better results. Until midi isn't rendered there's always the chance of something going wrong, glitches or any of those incoherent things that computers like to do every now and then. With audio I can really take care of the mix, fx, xfades, dynamics etc. From that moment on it feels like painting to me, choosing colors, refining things... not that I'm paricularly good at mixing, but that's how I like to do it.
This for me as well.

And? I sometime put FX on my VI tracks, then render with the FX, partially because it is the way I want it to sound, partially because then I can add different effects (i.e. blackhole to the midi to get a specific sound, but then a little room reverb to the audio to blend), but mostly because I've found sometimes I have latency problems that are easier to fix in audio and I also don't have 128 GB of RAM or a 16 core CPU.

I do keep the midi/instrument tracks but just disable and hide them to make it easier. Never delete anything. But I'm only a hobbyist and find as I get better, I will sometimes go back an fix things.
 
I've been discovering a good deal of timing issues with the playback on many midi VIs. Some worse than others. So I like to render them down to audio and correct any timing issues. It's a bit of a pain.

So everything is rendered to audio first. (Unfortunately).
 
The other good reason to render to audio is if you no longer have access to the VI, either because they quit supporting it or you rented it on something like EW Composer Cloud. I have a few tracks that were done by a former friend where the midi didn't get rendered. So now I have to figure out an instrument to use if I want to recreate it so I can fix some other stuff in the track to remix it. Long story that includes a dropped hard drive without a current back up.
 
Wonderful answers everyone, thanks a lot for clarification.

So, to recap:

1- I finish my music as usual in MIDI, edit, quantize, and all those midi stuff.
2- Bounce MIDI tracks in place to audio, edit, add cross fades, and all those audio stuff.
3- If needed, start to mix the audio tracks accordingly. I said "if needed" because I always see mixing as a reactive rather than a proactive approach. If it needs mixing, then do the mixing.
4- Keep the original MIDI tracks for later use indeed.

Is that all?
 
well I just want to reiterate, that you don't need to bounce to audio in order to mix the audio. That can still be the very last step before you archive the project frankly.

When you have midi tracks, they play through instrument plugins and the instrument plugins output audio that goes through audio channels in your DAW mixer where you can mix them exactly the same as is they were recorded audio regions. No difference.
 
well I just want to reiterate, that you don't need to bounce to audio in order to mix the audio. That can still be the very last step before you archive the project frankly.

When you have midi tracks, they play through instrument plugins and the instrument plugins output audio that goes through audio channels in your DAW mixer where you can mix them exactly the same as is they were recorded audio regions. No difference.

But isn't this still called MIDI mixing? I used to do this until I noticed many opinions on the web that it's better to bounce the MIDI into audio then mix because MIDI mixing introduce performance issues, especially when effects are applied to MIDI.
 
no

that is midi region playback through instrument plugins

vs

audio region playback directly

Either way could be MIXING through the audio mixer.

And yes, audio region playback uses less CPU. If you don't have enough cpu then you may be forced to bounce to disk earlier for that reason..but that is an entirely separate question then the question about whether to "mix" with midi or "mix" with audio.
 
no

that is midi region playback through instrument plugins

vs

audio region playback directly

Either way could be MIXING through the audio mixer.

Ahh OK, got it.

..but that is an entirely separate question then the question about whether to "mix" with midi or "mix" with audio.

Yes, I think I did ask the question in a wrong way because I am not expert so naive questions sometimes help me to understand things more. Sorry for that :sad:
 
I typically print stems then mix... I approach the mix with a fresh set of ears, there's no worries about defunct plugins if I need to revisit the project several years down the road... If I need access to the midi that's what the original project's for... Lots of benefits to working this way.

That said it depends on the project, sometimes you just don't have the luxury of time..
 
+1 to all the comments so far... I think even if you never render your audio tracks for whatever the reason, mix the audio through the mixer faders, not CC7. Partly for some of the reasons already mentioned, but a few more things to consider:

  1. Mixer faders have much more resolution then 0-127 as in the case of CC7

  2. CC7 is basically turning the volume of your instrument up and down. That effectively reduces the dynamic range of the instrument! You really want your instrument pretty much always set to CC7 volume all the way up...so that the full dynamic range of your instrument is available as the instrument intends it.

  3. Use CC11, Velocity crossfading, etc..to achieve instrument performance dynamics depending on how the instrument itself is programmed, but avoid using CC7 because of point#2 above.

  4. If you need to adjust the actual mix itself in some way..then use the mixer faders....higher resolution and it retains dynamic range of the instrument itself, the level is being adjusted by the DAW mixer.
Regardless of whether you're playing back instrument tracks, or playing back a recorded audio track, the audio ends up in the DAW's mixer as audio. Mix it all there.
I agree with you about Point #2, but JunkieXL said he sets his CC7 to 90 so he can adjust volume when needed if he needs a little more in the mix from something, this made me think 🤔 💭 is he onto something?
 
If you copy everything JXL does then you better set it to 90!

Honestly though I don't see the point of that. Why not just bump up the audio mixer fader when you need a little more?
 
The blanket statement that it's better to render to audio before mixing is nonsensical. Here's just one example of why.

Recently I was mixing a track that had a long sustained note that I decided I wanted to duck at the very end of the measure and let the long Reverb tail continue. It's a common effect.

It turns out though, that the reverb effect was built into the patch. If I had ducked this long note with the fader, or cc7 for that matter, it would have ducked the reverb tail as well.

But by using cc11 for said ducking, it just ducked the note and left the reverb tail intact.

Now I could create a long list of examples where it's better to go ahead and print or render a midi track. Taming unruly round robins, playback issues, CPU overload, Etc.

But this is a 'cake and eat it too' world. And you keeping your options open has no penalty.

As Long as you follow Beat's advice and make sure you render everything before you've completed the project, for posterity sake, then the only reason to choose one way or the other, is whatever makes your life easier or allows you to achieve your desired results.

There are no rules.

And all of the reasons I've seen presented here, as a blanket principal, for why it's better to mix with audio, just don't hold up to scrutiny. In fact, I would say 99% of the time, it just doesn't matter either way. You can open a project with fresh ears with live VSTi's. With the exception of crossfades or actually editing the audio waveform, you can pretty much do anything to a vsti that you can do to an audio track.

On the other hand, you can easily render in place a midi track, then edit the midi, and then render in place again.

It just doesn't matter.

But if you render out stems and use those to mix, you've just stripped away a lot of options unnecessarily.

But if you didn't need those options, then that doesn't matter either. It just doesn't matter.

On a side note, sort of, the reason why I like to mix using CC data velocity CC 11 is because years ago I became enamored with Old School Big Band performances where the musicians actually had to mix themselves on stage. This is a lost art form.

But I like it, so if I have the choice of using a fader to turn up or turn down a piano, or using velocity data, I'm inclined to use the latter.

Why? Because I like it. It's not better or worse. It's an artistic decision.

I also don't like the sound, in dynamic passages, of the room sound being turned up and down. So if I need to duck or modulate the dynamics of a string patch, for example, I think it sounds more natural to do it with CC 11 then the master fader.

But that's just me. I just want everyone to be happy. And more artistic choices, means more happiness.
 
Last edited:
I have to disagree with your statement that people here said there was only one way to mix.

Unless, of course, your contract requires a specific end product, which might include specific types of stems.
 
Top Bottom