#scoring stage

"Frankenstage" is what I call a small project of mine that I have been working on lately and which was the reason behind the comparison of microphone positions. My intention have been to try to match those microphone positions and emulate that soundscape created on the scoring stage, so that otherwise dry sampled instruments easily can be blended into the mix with instruments actually recorded at a scoring stage. The goal is not to recreate any specific scoring stage but instead create something inspired by the sound of stages such as MGM and Fox.

What I did was to use the different perspectives offered by CineSamples' sample libraries recorded at MGM Scoring Stage as ground truth and similar dry sampled instruments as reference. I then proceeded doing A/B testing of the signals from the ground truth and the reference with added processing. Most of the testing was performed by ear, simply listening to the differences and trying to match them using headphones, primarily listening to the coloration of the instruments and the color of the reverberation decay while taking into account that each instrument wasn't necessarily a perfect match in performance.

While the processing of the dry signal can be done in several ways with different effects, the way I elected to set it up was guided by my intention of creating something that could easily be duplicated by others. Thus avoiding any specific plugins and by using a freely available resource.

My tool of choice thus became the impulse responses from Samplicity's Bricasti M7 Impulse Response Library that can be downloaded from the website and used in a convolution reverb of your choice. The Bricasti M7 is an algorithmic hardware reverb by Bricasti Design, but unlike previous algorithmic reverbs it model natural acoustics and tries to produce as good early reflections as a convolution reverb that is using impulse responses from real places. Peter Emanuel Roos of Samplicity has created a library of TrueStereo impulse responses using this quite expensive hardware unit and have made it freely available on Bricasti's request. Note that you need a software convolution reverb capable of producing TrueStereo (or ganging two instances together) with zero latency to use it successfully.

I selected two impulse responses from the library after some listening, one modeling the early reflections and another to model the reverberation tail of the room. Since scoring stages such as MGM, Fox and Warner are treated acoustically with wood, I found the Medium & Dark ambiance preset being a good match for early reflections while the Large Wooden Room preset has a similar decay and sound as the reverberation tail.

My set up consists of three separate instances of these impulse responses, one for each microphone position, in a special split sends arrangement. The placement of instruments are thus achieved solely by panning and balancing the relative levels sent to each effect instance while no dry signal is sent at all to the master bus. By using this split sends arrangement I get the best things from both the send and insert effect world. The three instances are shared among all the instruments that need to be processed, just like any regular send effect, but at the same time the effect acts as an insert as it completely replaces the dry signal. It is similar to using a pre-fader send with the fader all the way down, but still being able to use the fader to balance the instrument relative to the other instruments as usual.

I found that the stereo image of the input in the reverb needed to be slightly tightened as the TrueStereo treatment of the impulse responses will widen it, which makes placement by panning easier. I also found that the amount of panning usually need to be slightly different between the signals sent to the different instances. The close signal need to be panned the most, the room signal (emulating the Decca Tree) a little bit less (about half as much) and the surround the least (again about as half as much as the room). The way I balance the sends is to first set the dominant perspective (usually the room) to 0dB, I then set the other perspectives to lower levels (usually between -6dB and -9dB for the close) until it sounds about right. If an instrument need to sound further away I just insert a low pass filter in the signal chain before the sends and attenuate the high frequencies. 

The only thing I am still not entirely happy with in this setup is the surround perspective as it does not yet sound wide enough compared to the ground truth.