Tuesday, June 28, 2011

Mix it till you puke

Dear readers,

This is a live blog update on mixing a song that is codenamed "Indian".
Edit: This song got its release name - Tale of the Liars.

Step 0: Record and fix any problems.
In the tracks we had, I noticed that there was no sense of space in many places, especially the verses, bridge and solo. This fixture was easy. Overlaying tracks with feedback and swell based guitar layers helps. The stereo swell/flange/delay effects consume the space but the main guitar doesn't allow this guitar to be prominent unless someone is listening deliberately to peel off the layers of production.

The heavy riffs were played sloppily at places in the original recording. Matching patches is a difficult task. Although the equipment is the same, the sound card and A/D converter is not. The trick is to get an approximately identical sound and then use a gentle multiband compression to iron out any differences.

Step 1: Export all chunks as consolidated stems.
It is far easier to work with consolidated stems than a jigsaw puzzle of multiple tracks that look like patchwork. Most big-shot engineers have assistants and interns who do this for them. REAPER has an option to export to stems. Some sections can be easily exported in that fashion. For the rest, it is an act of mute/solo and export as a render. At this point the folder has 392 files and working with such humongous data is inhumane. The task is relatively simple but is time consuming.

Set the output to 0 db or to a level below that so that individual tracks don't clip. Export each instrument/layer/stem individually as mono (preferably, wherever possible) or stereo files (where inevitable) so that when time aligned, they will play in perfect sync. Also, the number of files and tracks will decrease. When in doubt whether to print the effects or not into the stems, repeat the process twice - once with the FX and once with FX bypassed.

After the process is over, I am left with 28 tracks in 28 files, 2 of which are "doubt prints" with FX. This is a far better number than 392. It's manageable, its comprehensive and above all - it's filtered knowledge, not noise laden information.

Step 2: Import all chunks and clean up.
Most tracks will have long stretches of silences. These must be cleaned to reduce hard disk usage. Also grouping the members makes track management easier. For this project I have grouped them as the people who played them. Also, I made three aux tracks related to ambiance - a long plate, a short plate and a BPM synced delay. The master track also has a visualization insert - Voxengo SPAN. This one is an indispensable tool.

This step is also the very starting point of mixing.

Step 3: Get the drum, bass and vocals balance.
This is a preliminary approach. The levels will no doubt change, the EQs will get tweaked but the first thing is to get the drum balanced and then feed in the bass. Then I like to bring in the vocals. The guitars often work around this space.

What eventually happens is that I get a relative tonality of elements in the mix that work together, but in the end the levels are all over the place. However, as a whole, the drum and bass still stand well (maybe a bit buried or a bit OTT) in the mix. Earlier, I used to mix the instrumental and bring in the vocals. In this approach, the vocals feel like an outside element. In the current approach of building the tonal instrumentation around vocals, the instruments can be tamed. Trust me when I say this, taming instrument is a far better idea than taming and killing the vibe of the vocals.

The BPM synced delay (earlier step) is engaged to create a sense of open space in the solos and an interesting slap-feedback effect in a particular section of the background vocals.

Step 4: Do nothing.
At this point the vibe of the song starts to build and you'll have a fair a idea what is working and what is not. Time to give the song few good listens and make mental notes. It is better not to touch anything. A fresh set of ears is needed to move forward. Also, time to grab a CD for referencing.

There are few specific things I have to address in the mix I am working. Certain volume of an instrument fits well in a section while it sounds so OTT in another. Time for splicing and creating approximately (minor difference) identical lanes. This is not the time for automation and riding the faders.

Step 5: Mix, re-mix, mix at lower volumes, mix loud, re-mix at lower volumes.
This is the main time to follow the vibe and guts. Pull out anything, absolutely anything out of your arsenal. But remember one rule. No automation, no riding the fader. If I start riding the faders now, I'll be in a mess of a situation with tons of automation lanes and there will be nothing left that can be called as "fine tuning". So far the general mix sounds how I approximately wanted it to be. There are a few places I want the vocals to be a bit louder than other places, but that is reserved for a later time.

Occasionally mixing quite is something I do post step 4. This is something that I picked up from Andy Wallace. It helps your ears judge relative volumes better. However, one must also have fresh ears before attempting to move from louder to quieter volumes.

Step 6: Shape up the track using references.
There are two references. One is human (band/client,etc.) the other is the reference CD you chose. In my case, since my bandmates are the ones who'll co-approve the track, I send a streak of mixes with minor modifications, ride the fader and adjust the mix to the comments they send me as feedback. Also, constant referencing with the recorded material (CD) keeps the overall tonality in check. It took me 3 versions to get to a point where nothing major can be done (unless you take a 180 degree turn and undo the whole mix - in that case go back to step 2).

At this point the overall tonality is established and a coherent song is obtained. The volume will be low (since it is unmastered and you don't want clipping).

A small trick that I picked up from George Martin's interview - when the system is overloaded, it is better to bounce stems and relieve the system off its complexity. That's how The Beatles got such textured multi-layered sound on 4 track recording machines. The same applies when the FX overloads.

Step 7: Prepare for mastering.
Leave some decent headroom and make sure that the mix is not tamed with some brickwall limiter. Charles Dye and Brad Blackwood have some wonderful tips to share.

Since we cannot afford a mastering engineer (ME), I have to "master" it (if it can be called mastering). I like to keep a replay gain of -5 to -6 dB with peak just falling short of 1.0000 (float). The modern records have an RG of around -10 to -14 dB. These are fatigue inducing and will kill the eardrums, not to mention the dynamics and transients. My advice is to take it to a genuine ME. A good ME can do wonders with a well mixed track.