top of page

BA4 Creative Honours Project overview

With the view of trying to create sustainable employment opportunities as a composer, creative artist, and sound designer, this project consists of two main component parts. The first of these is the design and creation of virtual/sample instruments and the second being the creation of original sound and music compositions suited for a variety of uses including tv and film, video games, instrument demos, sync libraries and other bespoke commissions.

 

The music will be available through my website in addition to most of the common streaming platforms where the sample instruments will be shared through a platform called pianobook which is an online platform with a global audience which provides primarily free instruments for composers and musicians to use in their own work royalty free. The site is set up so that end users can tip instrument builders for their work and more recently a handful of samplists have been given the opportunity to work on paid for instruments which have been sold through Pianobooks parent company Spitfire Audio.

 

In addition to making use of the reach of this already established network, I plan to have these instruments as well others I am working on included on my own website as well as through Decent Samples which is a smaller online store, which allows small sample instrument builders to focus on creating instruments while they handle hosting and maintaining the platform in exchange for a commission (Hilowitz, 2022).

I will now provide a short overview of the instrument making process. Although there are many ways to achieve a goal, in my efforts I have attempted to be as systematic and repetitive in the methods employed  so that I may document my process and where possible make iterative improvements to either the overall process or any of the tools or practices I have developed in doing so. 

Key Stages

Sample instrument Building

Below is an early recording template for the 6 String Bass instrument I’m currently developing. It contains audio instructions of what notes to play and when, a click track, reference notes to assist with pitch accuracy, and lastly blank files which can be dragged onto the recorded audio tracks and will split the audio into separate sample regions (I have used this and a 3rd party method which I will go into more detail in another section). 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Before creating a sampling template like the one above for a more detailed instrument (multiple round robins and dynamic layers), it is necessary to ascertain how long each note will naturally ring out and this time, will then be the amount of time for each sample. Although it can be beneficial to follow the principal that the higher frequency notes will naturally decay sooner than the lower ones and adjust for each note saving both time and data storage. In this instance, I found that it working across multiple instruments each with varying characteristics, it was quicker for me to create a template where each of the notes were the same length from their lowest to their highest sacrificing overall time and data but allowing more uniformity at a higher quality across repeated measures. 

 

The next thing is to decide how many round robins (repeated notes) and dynamic layers the instrument will have as these each will help to convey a sense of realism. The trade-off is between perceived realism and the final data size as well as the amount of time and work required to develop a deeper sampled instrument.  

For this instrument I chose to have four round robins for the long notes and eight round robins for the short notes with two dynamic layers across each articulation. As I wanted this to be a very detailed deep sample instrument, I have also chosen to sample every minor third across the entire range of the bass. The following table should provide an insight into the amount of work an instrument of this detail will entail. 

 

Editing 

Once the recording and splitting of each of the samples has been done, it is necessary to edit some of the samples. Typically, this will be shifting sample start times, adding short fades at the beginning of each sample so to avoid clicking which can occur if there is audio across the zero crossing line or even shifting the start/end times if the cutters didn’t get it quite right. There are ways to batch process this, but it is always possible that the algorithm used will miss something so in most cases each sample will have been checked at least once. For a large sample set such as the 6-string bass this part of the process can be very time-consuming.

 

Since beginning this project I have discovered a set of tools, one of which is capable of splitting monolith files into individual samples as well as naming and sorting each one. 

 

Tuning 

After editing, it is necessary to tune some samples, this won’t always be necessary but for some instruments, this stage is an absolute must. The organ I am working on was in places as much as 40 cents flat on some notes and differing amounts sharp on others. For this process I use a combination of Melodyne and any other tools that can achieve the desired result with as few artefacts being introduced as possible. 

 

Noise Reduction

This is another essential part of the overall process. If an instrument was built using only one sample, it may sound fine on its own. But as notes are held down together or begin to overlap any noise that may present in the audio signal is magnified by the number of audio signals and the number notes being played together. 

 

Above is an example of the initial demo sample set I recorded for the bass.

The noise reduction is achieved using a program called izotope RX by analysing recorded ‘silence’ which helps to identify any background or signal noise and then setting the parameters to reduce/remove any undesired material. This process does remove some of the good frequency content form a signal if used too severely so there is usually a trade-off between noise and character.

 

Naming/Sorting

The next task is possibly the most labour intensive and that is to manually name and organise each sample by hand. I haven’t yet found a reliable consistent way of batch processing this task. However, I am currently revising my template to include a new audio track for every permutation of note (24 C0 Soft Layer RR1, 24 C0 Soft Layer RR2 and so on). There is then a feature in Logic Pro which allows the user to name audio regions by the name of the track. This would require some time to set up at the beginning of a sampling session, but I think could save lots of time manually editing every sample later. 


<Insert bit about new program>        

   

Coding

This has been the most difficult skill for me to acquire since beginning this project. I only learned how to code using the KSP (Kontakt) syntax halfway through BA3. When I began this project however, I knew accessibility was to be a focus and so elected to use Decent Sampler, which is completely free making it more accessible to the end user. Although I had only planned to create in Decent Sampler format, I feel like I may exclude a potential audience and have also made a counterpart in the Kontakt format which is an industry standard. 

 

Although there are some similarities in the types of functions required the syntax for each engine is quite distinct. Since the beginning of first semester, I have learned/taught myself by reading through the extensive developer notes, and now have a rough working code for every instrument I will be submitting as part of my project. I have deliberately chosen to highlight each of the features that the Decent Sampler engine is capable of (working within it's limitations as Kontakt has been development for much longer) whilst still realising very detailed, complex, and original instruments. It has been quite challenging learning instrument building and in two syntax' simultaneously has left me many times asking myself why? Despite the many hours just staring until I finally notice that one tiny error or its solution, the second the first sounds come out of the speakers is the most rewarding experience and to think that I can share that feeling (I hope) with others is a very special thing indeed.

Below are a series of screenshots of one of the smaller instruments (in terms of workload and final data size) prepared for the DS format to give an indication of the structure and layout required to create such an instrument. 

 

 

And here is the same instrument as coded using Kontakt KSP

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

GUI Design

For some of the background images I have used photos of the instruments that I have sampled but for others which are not strict facsimiles of the original instruments I have chosen to use AI software such as Dall E to create background image, I have also used photoshop to create my own as well. Regardless of where the image originates, I use photoshop to make sure they are scaled to the appropriate dimensions of the sample engine (Decent Sampler uses width="812" height="375" where Kontakt uses width=”633” height=”280”). In the below example I added the instrument and knob/slider labels to keep consistency between the DS and Kontakt versions.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

<Insert portion about new GUI design giving uniformity and cutting down on the work I would have to do. basically adjusting parameters to work within my scholastic timetable>

 

Beta Model

After I have finished working on an instrument, I have it checked by a few friends and colleagues as well as trying them on as many different DAWs, platforms, and operating systems. This stage is important as it allows me to catch any errors that may occur or more likely, things that I have already missed. It also makes sure there is some degree of consistency between users. 

 

Revisions

Below is a screenshot taken whilst working in Decent Sampler stand-alone plugin whilst editing the code to affect some changes that were identified during the beta testing.

 

 

 

 

 

Reflexive & Iterative Process

The completion of a Sample Instrument of any size requires many stages and a variety of different skills and hats. Each of the individual component parts will be continually revised and improved upon throughout the process. Since beginning this project, I happened to find a small one-person company based in Germany who has made several tools specifically for the benefit of sample instrument developers. They each have a specific function (splitting monolith files (a single recording which contains multiple samples) batch naming, editing start/end/loop points, etc) and used in conjunction, they each share workflow and data. This has made a huge difference to my overall efficiency and has allowed me to spend more time in other stages of building. 

Musical Creation

My musical workflow varies depending on the compositional concept and instrumentation. I use a blend of Logic Pro, VCV Rack (a virtual euro rack software), a hardware euro rack set up as well as variety of instruments, some of which will be played live, while some will be sampled and then triggered in midi from either the DAW or from my modular rig. 

 

Whether I am composing a cue or writing my own music I generally work from a template I have created inside Logic Pro. This has different folders for each type of instruments and already contains all the FX, buses, and routing that I typically use for each. Every instrument folder contains a selection of sample instruments pre-loaded as well as blank audio tracks. This helps me to stay focussed while I’m in a creative mindset so that I don’t waste precious time or energy looking for sounds or setting up basic things that I do in every session. It also helps me to be more consistent in both quality and aesthetically when working on multiple projects simultaneously. 

 

Although having the template set up this way allows me to mix as I go, I will begin a dedicated mixing stage once I know that I am finished creating/arranging the piece. I find keeping the stages separate and distinct, I can make more objective decisions with regards to the sound, space, and frequency structure of the piece. 

 

When I am creating in VCV Rack or directly from the modular, I will follow the same process and get to a place where I feel the creation/arrangement part is completed and I will then bounce the audio out from VCV and then bring it into Logic Pro for mixing. 

 

Timetable for completion/release

<Insert projected release dates table >

 

Bibliography

Hilowitz, D. (2022) Decent sampler developer resources - decent: Samples, decentsamples. Available at: https://www.decentsamples.com/decent-sampler-developer-resources/ (Accessed: January 27, 2023). 

Screenshot 2023-01-28 at 19.57.21.png
Screenshot 2023-12-13 at 11.53.31.png

Later Iteration able to calculate time required as well as data storage requirements at varying sample rates

Screenshot 2023-12-13 at 13.02.13.png
Screenshot 2023-12-13 at 12.57.48.png
Screenshot 2023-12-13 at 12.57.14.png
Screenshot 2023-12-13 at 13.04.36.png
Screenshot 2023-12-13 at 13.03.25.png
Screenshot 2023-12-13 at 12.58.53.png
Upright Piano Background.png
2 Voice Accordion Background.png
Background.png

Noise reduction requires a fine balance between removal of noise without changing the tone of the instrument.

The various GUI iterations of Magnus' Cavendish. 

A synth type instrument based on the Soma Lyra.

A test instrument I made to better understand the Decent Sampler group structure. Also used DallE to create a nightmarish background image.

Something coming next year.

The consistency of knob and slider placement saves a lot of time and energy that I can put elsewhere in the development.

bottom of page