Flux Alchemist

Whether you're rescuing an over‑compressed mix or extracting maximum punch from a drum track, Flux's innovative multi‑band dynamics tool could be just what you need.

The philosopher's stone is a legendary alchemical substance said to be capable of turning lead into gold. Alchemists spent their whole lives searching for it, and the 'science' of alchemy is thousands of years old. In naming their latest plug‑in Alchemist, Flux have signalled their intention to make a powerful tool that can effect a similar transformation with audio recordings.

More prosaically, Alchemist is a multi‑band dynamics tool, offering four different dynamic processes for each band — compression, de‑compression, expansion and de‑expansion — plus transient shaping, M/S width control and selectable level‑detection modes. Each band also has independent envelope controls such as attack, release and hold, and its own 'dynamic detection profile'. All these features make the parameter count skyrocket: the Pro Tools plug‑in automation panel reports a whopping 410 different parameters! Alchemist is available on Mac and PC in all the major native formats, and is authorised to an iLok or a Flux dongle.

Choices, Choices

Describing all those parameters in detail would take all afternoon, but in short, Alchemist can provide up to five bands of dynamics processing, applied to up to eight channels of surround sound. Each band has its own controls for input and output gain, filter frequencies, filter slopes, level detection mode (feed‑forward and feedback modes, with different integration time for RMS detection), envelope controls, dynamic profile, RMS/peak sensitivity, ratio, compression range, knee, M/S width, controls for transient shaping, Angel's Share and Hysteresis. The two last parameters are unique to Flux's algorithms, and were first seen in their Solera plug‑in, reviewed in SOS July 2008 (/sos/jul08/articles/flux.htm). Increasing the amount of Angel's Share produces a slightly less compressed and more open sound by relaxing the compression ratio applied when the dynamic range rises. Hysteresis makes the dynamic process less level‑dependent and instead focuses on the dynamic variations. It's a very cool feature, making it possible to apply the same amount of dynamic processing on drums even if they differ in intensity between verses and choruses. It also makes it possible to bring down hard consonants or nasal vowels in a vocal recording even if the vocal performance differs in level intensity.

Apart from compression, it's possible to apply de‑compression (adding dynamic range), expansion (expanding or gating away the low‑level content) and de‑expanding, which raises the low‑level content. De‑expansion is very useful for adding energy and excitement without smashing the transients, and it's one of the key features in Alchemist. It's also possible to add dynamics by applying transient shaping in the form of Bitter/Sweet, Flux's own transient shaping tool. Finally, the processed sound can be mixed with the original sound, and there's a useful clipper on the output that effectively takes care of any overshooting transients.

The amount of in‑band processing being done is shown by three vertical meters representing the envelope (equivalent to the gain‑reduction meter on a compressor), along with the dynamic and level differences between input and output. Below these, there's a horizontal graph showing the dynamic activity, which is helpful when trimming the Hysteresis threshold, and the graph under it indicates variations in the release time when auto release is engaged. The main graphical display contains settings for the filter frequencies and steepness of the filter slopes, and the number of bands; by right‑clicking on the display, you can copy settings from one band to another, or link them all together. By default, the dynamic activity of each filter band is shown in the display, and there's an auto‑scaling spectrum analyser showing the output signal. Finally, there's a display showing the resulting transfer curve of the dynamic processing in the selected filter band.

Single‑banded

With so many processes at work, Alchemist's display has a lot of information to feed back to the user!

To familiarise myself with Alchemist, I started by engaging single‑band processing on some well‑recorded drums. In single‑band mode, Alchemist offers the same kind of processing that can be done with the Flux Solera plug‑in. The default detection mode is named Solera and it seems to have the most responsive and accurate detection, but for drum‑bus duties I liked the feedback modes, because they produce a nice smack; even with heavy compression, the result is usable. Altering the knee setting makes the compression less obvious, and by tweaking the dynamic profile and auto‑release boundaries, you can achieve even smoother compression. The Hysteresis setting came into its own when dealing with a rather bass‑heavy floor tom: by increasing the hysteresis, I was able to apply pretty much the same compression to the kick/snare sections as to the kick/tom sections without altering any settings. Nifty. The output clipper also turns out to be really handy, because it shaves off overshooting transients without producing massive distortion. I managed to clip the output by around 6‑7dB without destroying the impact of the drums too much.

The opposite of compression is de‑compression, and when I engaged it in Alchemist, I was able to add punch and a little bit of excitement to the kick, snare and toms by adjusting the ratio and threshold. Altering Angel's Share and Hysteresis settings made the de‑compression more musical and less jumpy, and I was able to achieve better results than I've ever had before with de‑compression. Using the Bitter/Sweet transient shaping section is another way of adding or subtracting transients, but without being level‑dependent. The amount parameter goes from ‑100 to +100 percent, and it's possible to add Bitter/Sweet before or after the rest of the dynamics processing. By default, it uses the main stereo signal for processing, but it's also possible to engage an internal M/S encoder and shape the transients of the Mid or Sides signal before the result is decoded back to stereo. Adding transients to the Sides signal can open up a drum track without over‑emphasising the impact of the drums in the middle. Bitter/Sweet is snappier and harder‑sounding than de‑compression, and I found it generally more useful.

Multi‑band Chemistry

Switching to multi‑band processing unleashes the true power of Alchemist, as I discovered when I gave it a run on an orchestral piece intended for a computer game. The mix was rather dense, so I dialled in a four‑band process with both compression and de‑expansion. A normal multi‑band compressor would not have been able to handle the macro‑dynamics of the orchestra playing the intro softly, followed by the main theme at full force, but when I increased the Hysteresis parameter, Alchemist handled both parts without putting the lid on the stronger theme part. Compressing the low end worked remarkably well, and with added Hysteresis, the subsonic booms were controlled, but without taking away their impact. Adding some de‑expansion made the softer parts more exciting and energetic but also brought up some reverberation. The Bitter/Sweet transient shaper was very seductive, enhancing the transient definition of each filter band and creating a revitalising effect, but the effect was easily overdone, leading to a less natural result.

Zero Latency

In my view, most multi‑band compressors have two flaws. First, the filter slopes are too steep and the compression knees too hard, which can make the processing audible and far from natural. Second, they use linear‑phase filtering to separate the bands. In theory, this is the 'best' approach, as there's no phase shifting between the bands, and the filter slopes can be made very steep, but it introduces transient pre‑ringing and latency. Interestingly, the filter slopes in Alchemist can be adjusted between extremes of 18dB and 54dB per octave, and the filters themselves aren't linear‑phase. This means that the default process latency is zero, which I think is great, because plug‑in latency has a tendency to stack up in Pro Tools and eventually stab you in the back. If it's vital that the dynamic processing react really fast, lookahead processing delay can be added to each filter band individually, allowing the attack time to be set to zero, if need be.

Against The Competition

My main multi‑band compressor is the TC Electronic MD3 plug‑in, which always seems to produce smooth and musical compression no matter what I throw at it. Comparing Alchemist to MD3 made me realise how gentle the filter slopes are in the latter. This clearly made it unnecessary for TC to use linear‑phase filters, and also means that the bands blend and interact with each other much more than in other multi‑band devices. The compression knee is also quite gentle, which makes the processing less intrusive. Just for the fun of it, I tried to match the performance of MD3 with Alchemist by setting the filter slopes to 18dB/octave and making the knee values quite gentle. When I set all the parameters to the same values as in MD3, with fixed release and feed‑forward detection, and matched the amount of compression by ear, Alchemist produced a slightly flat‑sounding result with less depth. But then I decided to keep the same attack, ratio and knee settings but change the detection mode to Classic Feedback Fast and enable the advanced release mode, which adapts the release time to the material. I also add a bit of Angel's Share, which eases off the compression in a nice way. Now the differences became much smaller, and Alchemist produced an impressive sound with some extra punch. MD3 still has a roundness that I've grown to love, but Alchemist is not far behind, and I'd say it's a case of apples or oranges. Alchemist is infinitely more flexible, and doesn't need any dedicated DSP hardware to run. I think it's a great companion to TC MD3, and both have their unique qualities.

A favourite dynamic filter of mine is the excellent Sonnox SuprEsser, which I use for taming harsh-sounding vocals by dynamically suppressing the area between 2000 and 4500 Hz. SuprEsser has an adaptable threshold and handles troublesome frequency build‑up even if the dynamics change. I tried to replicate the process with Alchemist on some nuked rock vocals, and it took a bit of tweaking because some of the parameters interact, but in the end I got great results. SuprEsser is slightly smoother‑sounding, but Alchemist keeps a little more 'bite', and I'd be perfectly happy using either of them in a mixing situation.

A Kind Of Magic

It feels like I've only scratched the surface of what Flux's Alchemist can do, because it's truly a remarkable dynamics tool, unlike any other plug‑in. Packing all kinds of dynamic processing into one plug‑in creates a lot of parameters, and I can't help but think that there must be a way to design a less cluttered layout — selectable easy/advanced views to hide or show parameters would one way to go. Flux should also consider adding a lot of factory presets, because they are helpful for showing what it's capable of: right now it comes with just 10 presets. I also miss the possibility of full M/S encoded processing, because it's handy in mastering situations.

Whether Alchemist will turn your mixes into gold still depends on your arranging and recording skills, but it's a very potent dynamics tool. There's a trial version on the Flux web site, so why not try some alchemical sorcery of your own?