My problem with sample libraries

→ in deutsch lesen

When it comes to producing music, I utilize various digital software tools to achieve an authentic sound. In the realm of digital sound generation, there has been, in my opinion, a tendency towards redundancy in recent years. I believe that too many resources are being invested in seemingly new but fundamentally repetitive products. In this post, I want to shed some light on this issue more specifically.

Premise

In order to avoid having to rent a real orchestra, record it, and edit it afterwards, I rely on a technique called sampling (see Wikipedia for reference). You can imagine it like this: Individual tones of an instrument have been recorded and programmed into a software tool, allowing me to play these tones with my keyboard using my own melodies and harmonies. Such products are called sample libraries. Some sample libraries have recorded entire phrases, while others have individual notes at different volumes. Economically speaking, this saves time and costs in many ways, as mentioned before, as there is no need for a real orchestra. However, it's important to consider that this technique has its limitations. Real recordings of instruments generally sound more authentic, at least to the trained ear, compared to the digital copy.

I must mention at this point that I am somewhat biased when it comes to sample libraries. Therefore, this is not a scientifically exhaustive post. However, I have tried to base this post on data as much as possible in order to validate my initial assumption in a more objective manner.

Conventional vs. innovative

I want to make a differentiation here. Conventional libraries, to me, are those that use static samples and rely on minimal scripting. Scripting refers to a form of programming that utilizes samples depending on the user's musical performance.

On the other hand, an innovative tool (I don't think "library" fits here anymore) primarily relies on scripting or programming. It could be a library that uses relatively few samples to generate a procedural sound. Alternatively, it could be a completely generated sound that attempts to simulate the physical characteristics of an acoustic instrument through algorithms. In this case, the static aspect of a conventional library theoretically disappears since the parameters of the algorithm are controllable.

Therefore, anything that primarily focuses on samples is considered conventional, while anything that centers around programming is deemed innovative. One possible way to measure this could be through the size of a product. However, it's also possible that a provider explicitly mentions the use of such technology, eliminating the need to even consider the size in order to recognize an innovative product.

Problems I suspect

My perception is that there seems to be an overwhelming number of conventional sample libraries flooding the market, particularly for string instruments. I am constantly browsing through various channels to stay informed about current tools for music production (german url, sorry). Every now and then, there are announcements about new conventional string libraries, among other things. However, I find this conventional approach increasingly inadequate, redundant, and outdated. The following points will attempt to explain why and what I specifically consider problematic.

Lack of flexibility

The premise is that the primary basis of such a product is the static samples. There may be libraries that have built a script around these samples to make the sound slightly more modulatable. However, in the end, the underlying fundamental sound generation still relies on the sample itself. As a result, in my experience, there are often very few parameters that can be controlled to significantly vary the sound. For different playing styles, you would primarily need different instruments in your music production software (or so-called keyswitches, which allow you to switch playing styles using a note outside the tonal range - but it's binary, meaning either on or off). To me, this seems like needing to have two different guitars on stage, one for playing chords and another for playing individual notes. Therefore, I see a lack of flexibility when it comes to the usability and sound of the product.

Lack of sonic innovation

The diversity of sound is also technically limited, as static recordings are repeatedly played back. The sound character of the library is mostly determined by factors such as the recording space, instrument positioning, microphones, post-processing, etc. Many composers use these static samples to create their music. Strictly speaking, it is possible that composers, who use the same sample libraries, may have a similar sound character (thus not being innovative in my opinion). In this regard, my other blog post might be interesting: How I Assess Music Quality. There, I relate innovation to music quality.

Lack of resource efficiency

I see difficulties both from an ecological perspective and an economic perspective. Conventional sample libraries are usually very large and require high bandwidth to download the product. On which servers are these products stored? What power source does the customer use while the PC is running to load the products? If the internet is slow, the factor of time also comes into play. Ultimately, libraries must be stored on the hard drive to be usable. While hard drive storage may not be too expensive anymore, it wouldn't be a compelling argument for me if it could be avoided. Furthermore, to utilize such libraries optimally and smoothly, they should be used on an SSD. That's where the cost can quickly increase if hard drive space becomes tight!

One could argue that innovative tools require computational power and thus continuous electricity instead of relying on hard drive and RAM space. New CPUs consume less power than older ones. Additionally, such innovative tools continue to evolve, and their programming (hopefully) gets continuously optimized.

Oversaturation

In addition to the previously discussed technical aspects, I also see a problem in the oversaturation of the market. As a composer, when I'm looking for new tools or specifically searching for a better string library, I easily become overwhelmed. The market, in my opinion, offers far too many conventional libraries. How can one navigate through this oversupply? The issue becomes even more problematic, since, in most cases, you can't test such a tool before making a purchase. So, you're essentially investing money based on trust in the sound examples provided by the vendors (unfortunately, this applies to innovative alternatives as well!).

Investigate assumptions with numbers

To verify the presumed issues with numbers, I attempted to research current data. I counted the number of providers, their string library products, and related figures such as the size and prices of these products. First, I visited KVR to search for the latest sample libraries for strings. In my experience, this website has always been a reliable source for up-to-date music tools. I was able to identify current providers through this platform. Additionally, I visited other developer sites that I have become familiar with over time.

As a reminder: I want to differentiate between conventional and innovative string libraries. The distinction can be found at the top of this post.

Table: conventional string libraries

All links in the table were visited on 12th September 2022. I have attempted to determine the number of products, required hard drive storage, and the price. Additionally, I have calculated the price per gigabyte of samples. Researching was sometimes difficult because the shops had varying and partially confusing layouts. Unfortunately, I had to exclude some products as it was not apparent how many string samples were included, and the price was not transparently listed. Furthermore, there are many more providers of string libraries. This is only a selection from the current libraries according to the KVR search query and from providers that came to mind spontaneously. The source link is always the developer's name. The price is converted to euros according to the exchange rate on that day.

|Provider|Products|Size|Price|GB/Product|€/Product|€/GB| |---|---|---|---|---|---| |Heavocity|1|38 GB|540 €|38 GB|540 €|14,21 €/GB| |Strezov Sampling|6|221 GB|1172 €|37 GB|195 €|5,30 €/GB| |Impact Soundworks|2|84 GB|470 €|42 GB|235 €|5,60 €/GB| |Native Instruments|9|296 GB|2491 €|33 GB|277 €|8,42 €/GB| |VSL 1, VSL 2|5?*|1441 GB|4335 €|288 GB|867 €|3,00 €/GB| |Orchestral Tools|7|686 GB|1499 €|98 GB|214 €|2,19 €/GB| |Cine Samples|7|>155 GB*|1814 €|22 GB|259 €|11,70 €/GB| |Spitfire Audio|>26*|>2641 GB*|9334 €*|102 GB|359 €|3,53 €/GB| |8Dio|34|1120 GB|5834 €|33 GB|172 €|5,21 €/GB| |Audiobro|2|170 GB|933 €|85 GB|467 €|5,49 €/GB| |Chris Hein|6|97 GB|1474 €|16 GB|246 €|15,20 €/GB| |Musical Sampling|3|12 GB|736 €|4 GB|245 €|61,33 €/GB| |Soundiron|2|23 GB|198 €|12 GB|99 €|8,61 €/GB| |Total|110|6984 GB|30830 €|||| |Average|8,46|537 GB|2372 €|62 GB|321 €|11,52 €/GB|

*The respective shop was not transparent enough to extract the necessary information. The number of products, the size of the product, or even the respective price were sometimes not provided. Therefore, the information provided here is kept vague.

Table: innovative string tools

All links in the table were visited on 12th September 2022. The data is the same as in the table above. Please note that there may be other providers of these tools that I am not aware of or simply couldn't find. The fact that I had difficulty finding more providers (if they exist) is a bad sign to me. It leads me to suspect that this technology is less widespread. The source link is again the developer's name, and the prices are converted to euros based on the exchange rate on that day.

|Provider|Products|Size|Price|GB/Product|€/Product|€/GB| |---|---|---|---|---|---| |Sample Modeling|1|5 GB|299 €|5 GB|299 €|60,00 €/GB| |Audio Modeling|4|1 GB|480 €|0.25 GB|120 €|480,00 €/GB| |Expressive E|3|0,05 GB|147 €|0.014 GB|49 €|2940,00 €/GB| |Synful Orchestra|1|0,16 GB*|472 EUR €|0.16 GB|472 €|2950,00 €/GB| |Aaron Venture**|1|? GB|393 €|? GB|393 €|? €/GB| |Total|10|6,21 GB|1791 €|||| |Average|2|1 GB|358 €|1 GB|267 €|1286,00 €/GB|

*I am familiar with the size of the product as I own it. There is no information about the size on the website.

**Unfortunately, the product is not yet available. Nevertheless, I wanted to list it to demonstrate that there is another provider with innovative technology.

Evaluation of the tables

I would like to evaluate the tables again in relation to the following questions regarding conventional string libraries:

  • Is there market oversaturation?
  • Is there a lack of resource-friendliness?
  • Is there a lack of sonic innovation?
  • Is there a lack of flexibility in product usage and sound?
  • How transparent is the market here?

For simplicity, I will refer to the conventional string library as Conventional (upper table) and the innovative string library as Innovative (lower table).

Analysis: oversaturation

When comparing both tables, one should already notice: there are 13 providers for Conventional and only 5 providers for Innovative. Among the latter mentioned providers, there is also one that has not been active for many years. Thus, according to the tables, one can say that there are essentially more than twice as many providers for Conventional compared to Innovative. My intuition tells me that further research would reveal an even more pronounced imbalance.

Now, if we simply count the products listed in the tables, we get 110 products for Conventional and 10 for Innovative. Here, there is 11 times more Conventional than Innovative. Even the consideration of products per provider shows an imbalance, with 8.46 to 2 products. In my opinion, this indicates an extreme oversupply of Conventional.

Evaluation: resource-friendliness

For resource-friendliness, let's first examine the digital file size of the products. Once again, the Conventional stands out with 6984 GB compared to the Innovative with only 6.81 GB. Even the GB size per product indicates that 62 GB are equivalent to a single gigabyte. If I were to measure resource-friendliness based solely on size, it would be evident that the conventional option is remarkably unfriendly in terms of resource utilization.

In fairness, I would like to mention that computational capacity, specifically CPU usage, also plays a role. Based on my own experience, resource usage in this terms tends to be significantly higher with Innovative. Unfortunately, I find it challenging to test this aspect thoroughly as I don't own all the products mentioned. I can only provide an extremely rough and spontaneous "benchmark":

I compare products I can test myself using my technology (i9 9900K). In this test, I have higher string players perform a staccato motif, monitoring the CPU usage displayed by my music software, Reaper.

Conventional:

Product CPU-usage
Musical Sampling Adventure Strings 3,6%
Audiobro LA Scoring Strings 2 Lite 2,7%
Average 3,15%

Innovative:

Product CPU-usage
Sample Modeling Strings 11,6%
Synful Orchestra 1,8%
Audio Modeling SWAM Viola 11,6%
Average 8,33%

If one were to trust this spontaneous mini-benchmark, it could be said that Innovative requires slightly more than twice the CPU usage. Accordingly, one can now consider how to weigh this. After all, I am comparing a 62-fold disk and bandwidth usage to a double CPU usage, at least from a technical standpoint.

My hope is that for Innovative, the programming will improve and CPUs will become more efficient. As for the data size of Conventional, I tend to observe that such products seem to be getting larger and larger.

Analysis: sonic innovation

Conventional libraries typically have a patch list (for example, Adventure Strings). This list specifies the playing techniques recorded for the product and playable by the user. Such lists are not necessary for innovative products, as these products allow for live playing techniques within the music software. For example, you can perform a rapid trill by playing quickly alternating notes, without relying on a pre-recorded trill. More on this topic under the flexibility section.

When examining patch lists in conventional libraries, they usually include "Legato / Sustain, Spiccato, Staccato," and sometimes "Pizzicato, Tremolo, Trills." These include long notes played legato, short notes played staccato, and sometimes plucked notes, among others. Without being able to directly compare all products, I still doubt that the legato from Product A sounds significantly different from the legato of Product B. Also, always offering the same playing techniques is, in my opinion, not very innovative. There are exceptions, such as CAGE Strings by 8Dio. This product doesn't even include legato or staccato patches. That seems more innovative to me. However, since this provider offers 34 products, it's still just a drop in the ocean for me.

Innovative libraries generally allow for much more flexibility, enabling you to achieve a more unique sound. For example, in Sample Modeling Strings, you can adjust many parameters to customize the sound of the instrument. Such flexibility would be difficult, if not impossible, with mere static samples.

I must also note that playing techniques like pizzicato in Sample Modeling Strings are also switched using so-called keyswitches. Thus, at this point, this library behaves almost like a conventional library.

Evaluation: flexibility

This brings us directly to the point of flexibility. The aforementioned customization options provided by Innovative are, in my experience, completely absent in Conventional products. However, the range of adjustment possibilities is not the same for all Innovative products. I recall (although not specifically related to strings; nevertheless, a good comparison in terms of the aspect) being able to modify values in WIVI Brass to make the instrument larger or smaller. In Aaron Venture Infinite Brass, on the other hand, such options are rather limited. Both products fall under the category of Innovative tools.

As mentioned earlier, there are also playing techniques themselves. In Conventional products, these techniques are recorded and playable. In Innovative ones, however, they can be directly performed within the music program. As a result, it is sometimes possible to vary the speed of a trill in Innovative products, whereas in Conventional ones, it is determined by the recorded performance. Additionally, in Innovative products, a vibrato can be adjusted variably. Conventional products can, at best, mix two separate samples back and forth.

Analysis: Transparency

Let's take a look at the first tables again. I once tried to calculate a price per gigabyte. These prices fluctuate greatly. For example, Spitfire offers a product at 1.71 €/GB. However, the same provider also has another product in its lineup priced at 19.15 €/GB. At 8Dio, the price even goes up to 77.82 €/GB. This means that within the Conventional range, we already have price differences of a factor of 45.

However, this factor is even greater for Innovative products, reaching nearly 50. The GB prices themselves are also enormous in comparison, sometimes going up to 2950 €/GB. However, this is due to the underlying technology. Because these products are primarily based on programming rather than samples. Thus their size typically comes from the programming itself. In contrast, Conventional products assign financial value to the sample itself and the size of the data.

Perhaps it would be conceivable to compare the CPU usage of Innovative products in the same way. According to my mini-benchmark, there is a range of 1.8% to 11.6% and therefore a factor of just over 6 in difference. However, this perspective seems increasingly strange to me. After all, Conventional products often advertise the incredible amount of data they offer. It's as if they want to reach as many people as possible by boasting large numbers in order to sell the product. It would be odd to advertise with high CPU requirements.

Conclusion

I admit that I approach this topic with some emotion. I assume that most providers primarily act with a profit-oriented mindset and have no real interest in bringing something new (i.e., something innovative) to the market. There is a new product that has so many patches or samples (in my opinion, it must always include "deep sampled" and a large gigabyte size). And then they claim to have "striking realism" or "true emotions" or be "next generation deep sampling". Personally, I just can no longer take this buzzword bingo seriously. I also wonder if the providers themselves truly believe they are creating something innovative, or if they simply don't care and want to join the buzzword bingo bandwagon. As a composer and a potential customer of theirs, I feel somewhat taken advantage of.

Furthermore, I don't believe that the future should reside in this landscape of products contaminated, in my opinion, by marketing techniques. I consider the techniques of conventional products already exhausted. The innovative approaches I have listed here should be more prevalent. Apart from that, I also find the "marketing presence" of the respective providers much less negative, if they even have such a presence. For example, videos by Aaron Venture usually show screen recordings of a production, demonstrating the sound and a potential project view (in the music software). Certainly, these providers have less money at their disposal, or, in my optimistic hope, the funds are being put to better use in innovation and technological research.

I find it extremely regrettable that Wallander Instruments (mentioned earlier only for comparison; the company unfortunately never had string instruments) and Synful are two companies that either no longer exist (Synful's copyright is stated as 2006), or have significantly changed their focus (Wallander now produces Noteperformer). These products were not only small (< 1 GB!) but also had very low CPU usage, making them resource-friendly overall. Of course, their sound was not as authentic as current products (e.g., Aaron Venture, Sample Modeling, Audio Modeling). However, I wonder what these products could have become.

Overall, I hope that the major providers of conventional products will eventually change course and venture into the realm of innovative products. I believe that the future should lie in these products because they offer composers more possibilities and flexibility. Moving away from static samples and towards more procedural sound technology.