MIDI vs. Live Recording: An Orchestration Exercise

Here is an arrangement and orchestration I did around seven months ago, of Vanessa Paradis’s Divine Idylle. The original recording was arranged in a 60’s-inspired style. For this project, we made new versions of her songs for the orchestra, which were really different from the original arrangements. (to listen to the original recording of Divine Idylle, view her music video in Youtube.)

MIDI VS. LIVE arrangement

Now, here are the two “realizations” of the arrangement I did for this song:

The first one is a MIDI version. The bulk of the samples are from the Vienna Symphonic Library, with a few Kirk Hunter Strings and some East West instruments. It doesn’t have vocals yet here... Also, I posted my draft mix, in order to illustrate some stuff that could be improved on (which I will go into detail later). Listen to this one first.

Divine Idylle MIDI version ©Carmel House Studios, 2009

(Pardon the copy-protect thingy, I had to insert annoying “reminders” throughout the mp3 clips so as to discourage unauthorized use)

The second one is a live version, recorded by FILharmoniKA at Carmel House Studios. I sang a demo vocal in it, in French. (I did my best, ok?)

Divine Idylle LIVE version ©Carmel House Studios, 2009

To be able to compare the two, side by side in your own DAW or sequencer: Set the tempo to 90 BPM, create an audio track for each mp3, then start them at bar 1. Just mute or solo one track if you’re listening to the other.


Now let’s look at how the MIDI version could’ve been improved on, based what we’ve heard from the live orchestra version. (Or if there are parts of the MIDI orchestration which I prefer over my live orchestration)

1. BAR 2 onwards: MIDI version has too much reverb on the snare drum
- this was pointed out to me in one of the orchestration forums I frequent.

2. BAR 2-10: I spent a lot of hours automating the MIDI string tracks in order to have them imitate the sound envelope of a large string section’s attack and decay (on slow-moving lines), and I thought I was able to make a convincing sound, till I heard the live version. I think strings are one of the most difficult instruments to realize in MIDI! Because even when you’ve got the sound envelope right, there are a vast number of other factors to consider: String players usually don’t perform their vibratos at the same time, and at the exact same speed, and there is some sort of “sympathetic vibration” going on.. their overall sound is just massive and lush.

A technique that a mentor of mine shared with me, which helps a lot in MIDI orchestration is this: DON’T quantize your arrangements, as live players don’t play at the exact same millisecond anyway.

3. BAR 2-5: MIDI version used an English Horn.. but an English Horn was not available during the live session, so I had a Clarinet substitute for the English Horn lines. I think it worked nicely, though.

4. BAR 6: The attack of the Trumpet Ensemble on the MIDI version could’ve been less biting.. the live trumpet players were able to adjust to the mellow treatment of the strings on that particular line.

5. BAR 9, live version: Wrong note on the brasses.. whoops!

6. BAR 23: Hmm, I don’t know whose sound I prefer over the other here, the MIDI trombone ensemble, or the live one (a matter of preference in mixing). I sometimes like that “brassy” sound to the trombones in that register (which I did in the MIDI), even in a ballad setting.

7. BAR 27: I revised some string lines for the live version here because in the MIDI version, they just seemed to “drift off”.

8. BAR 36: Again, something I find difficult with MIDI strings: attaining a soft and “round” sound. Too often, the string samples are too bright or too striking. For a soft passage, one has to play the samples softer, but the softer velocity layers in samples don’t have that intensity or life which live players bring to the session. I’m sure there are workarounds here (maybe use a small cello ensemble along with the viola part?). VSL already contains some of the “darker” string samples out there, but will still need a lot of tweaking to achieve a convincing mellow sound.

9. BAR 44-45: Okay, that MIDI oboe didn’t quite pass the test for “almost sounds live”. Notice how in the live oboe recording, there are small, almost unnoticeable “breaks” in the line, which occur during the player’s key-switching. In my effort to make my MIDI oboe sound legato, I forgot that small but important detail, thus making the MIDI oboe sound really MIDI, as if a keyboard was playing the notes (which is what we really want to avoid in MIDI orchestration). I wonder if using an Oboe Legato patch could’ve solved the problem?

10. BAR 46-49 Now here’s one part where I liked the MIDI strings, because they seemed to have life in them. But the line would’ve worked better if the string reverb was brought low, even if only in this part.

11. BAR 50-51 The woodwinds in the MIDI were too “up front” and too loud to be convincing.

12. BAR 52: Another instance where I prefer the mix of my brasses in the MIDI over the live one.. I would’ve liked the trumpets to “blare” a bit, especially in that register, but they were somehow drowned out in the live version. I wanted the brasses to be up front here because they were a “response” to the woodwinds in bars 50-51. Again, this is a matter of preference rather than a textbook rule.

13. BAR 55: In the MIDI version, you could still hear the woodwind lines, but notice how they were drowned out in the live version (or to put it better, they “blended” well with the other orchestra instruments that they doubled in unison). As an arranger, I would’ve liked the woodwind lines to have been heard, but in reality, they would’ve really been smothered by the rest of the orchestra in that register, also because of the dynamic level of that part of the piece. I’d admit that this is where my weakness in mixing lies.. in MIDI, I sometimes make the mistake of highlighting a certain passage in order to make it come out, even if by “live orchestration standards,” the passage could not have floated over the rest of the orchestra. (Unless, of course, we are talking about a soloist, in which case the instrument will really have its own separate mic).

14. BAR 78: I forgot to put the crescendo mark on the brasses (which was there in my MIDI), so they just sort of drifted off in the live version. But I think the “drift off” worked better.

15. BAR 81-84: Hmm. It’s only now that I realized I forgot to put all my cymbal washes into the live version. He he! Too bad, it would’ve really added to the effect I’ve intended.

- - - - - - - - -

There.. hope you liked this mini-analysis!


 Subscribe in a reader

Related posts:

Theme from Ipis Man
Things I Learned in a Forum