Now that I've completed my third "The Experience" piano for release for free in December (Pianobook), I've had some time to finally play with one of the VSL 30 day demo's. I thought I'd write some thoughts about it coming from a samplist's perspective, highlighting both VSL and my own approaches. @Del Vento documents his hybrid piano hijinks, so I'll contribute with some of my lessons learned here.
So far, I've only activated the Bluthner, and have yet to activate the others as I'm trying to extend my enjoyment of the demos. I realize that they are all different in sonic character, and also the others may have other features I haven't tried yet, like the sympathetic resonance.
My libraries and the VSL libraries really come from the opposite end of the spectrum. One could say that VSL has gone through extreme lengths to create the best representation possible. They have a custom recording hall, use high end mics, sample 4000+ samples per key, have their own sample player, prepare ultra high end pianos, etc. My journey has been scrappier, and has evolved from the 80/20 rule, at least my interpretation, which is that I can achieve a product 80% as good with 20% of the effort. And then continue to optimize by applying my resources to the most important parts of the problem where they are lacking. Incidentally, I think other commercial developers fit in somewhere in between and the flowery marketing-speak that I read on their websites actually is amusing/annoying to me as I can see through it.
Anyway, I started during the Covid stay at home period with my own piano, using some Audio Technica AT2035's and a Samson GoMic, which are basically a high quality low end condenser mic and a podcast mic for real beginners. I realized that mixing two mic perspective streams is much better than one. Not wanting to invest in $400-800 mid range mics x 4 or x8 for this hobby, I made my own mics using plans from the MicBuilders forums. I had not realized that i could build such high quailty mics myself:
I think I spent maybe $400 total on my custom mics, and I have 8 capsules that I can position in custom mic patterns, 4 of which are placed for VR/Ambisonics based on my experimentation. The entire VR/Ambisonics thing is an entirely different story. Let's just say that I'm ahead of my time in the way that Zuckerberg is head of the curve with the Metaverse.
Back to the comparison:
VSL bought these pianos (I think) for their studio and are available to rent for studio recordings. By most measures, they are the highest end concert grands that are available. For my libraries, I just asked friends and met knew ones in the Bay Area (near San Francisco) who had nice pianos. Granted, the VSL grands have better bass as they are 9+ feet long pianos. I'm still trying to get access to more concert grands like my Fazioli F308, but for now 7 foot semi-concert grands I think are pretty good. (incidentally, if anyone knows of someone with such an instrument in the Bay Area, let me know)
This is one area where I think my goals and approach differ from VSL and most commercial libraries. First of all, if I won't pay for microphones, I won't pay for studio or concert hall time. I want my pianos to feel like I'm playing in my living room, so I don't need to hear the reverb of a concert hall. And if I wanted to feel like playing in a concert hall, there are many convolution reverbs to use to get that feeling. VSL's bluthner, while being recorded in their smaller studio, still has 2+ second release triggers. For me this takes away from the realism of playing in my small living room. Great for recordings, and I'm sure the intent is to be able to mix with other concert instruments that have been performed in a larger surrounding (ie, an orchestra). I record as dryly and as close to the strings as possible to take this part out of the equation.
I also have to deal with a normal home's condition - people don't realize what "I need it quiet" really means. Birds outside, gardeners with blowers across the street, cars driving by, even the hum of a refrigerator cause me headaches. It means I need to re-record if I hear noise, and then spend hours post-processing later. Taking out noises with software is pretty cool, but can get really tedious.
Time spent recording:
VSL says that they spend weeks recording their pianos. I typically spend about 4-8 hours total, including setup, recording, and teardown. However, to not bother my friends too much, I do a pre setup on my own piano to calibrate mics and get the stand angles perfect before I move to their homes. This now takes me maybe 2-3 more hours.
VSL uses a crazy amount of samples and velocity layering for the industry's best (save PTQ maybe) playability. Each note (88 keys, more for the BI) is sampled in more layers than I can count.
I manually perform seven layers by looking at a VU meter, and I only record minor thirds up the piano (A, C, D#, F#). I use the sampler to pitch it down. I experimented alot with this at first, because Pianobook recommended beginners start with fifths, but major scales (let along chromatic scales) are really annoying when it's the same sound pitched down. To this day, I think this is the achilles heel of my libraries. Some chromatic passages are annoying - play C, B, A# repeatedly and you'd see what I mean. Flight of the Bumblebee is not fun to play. I think I've thought of a way to create more randomness with EQ here - so maybe later it won't be a problem.
I use EQ and amplification techniques to get the velocity layers to blend into each other. I manually play seven velocity layers for each note, which is about all I can do when I spend such little time in someone's home. It's amazing how well this works.
Below is the video that taught me this. They made a 2 layer piano library:
More layers and more samples are definitely better, but this shortcut makes a meager amount of layers feel real. This is also one reason my Kontakt instrument is better than the SFZ one. Optimizing the EQ in SFZ is more difficult and on the to-do list.
As said before, I want to be in the headspace of the piano, not in the audience. So my sample libraries have a wider image in general, and lack a focused image. VSL on the other hand, has close, mid and far microphone placements - and a large hall.
@Gamma1734 asked me how I could get that bright resonance out of the Yamaha. It's because of the damper release triggers. VSL calls it "Advanced Release Sample Technology". This is basically recording the decay of the resonance when the key is lifted and the dampers mute the strings. Put the mics inside the piano (as I do), and the mics record the sounds bouncing around inside it after the key is released. Took me awhile to figure this out using some examples from collaborator(s) here on PC. Reveal yourself if you like (!)
VSL has their own engine and they can allow the user to control the trigger points. I'm limited by Kontakt and SFZ engines, and opted to just have the feature with only 3 range options (in Kontakt). SFZ doesn't have a GUI, so it's only one pre-set. The Kontakt engine isn't great for half pedaling. I won't support DecentSampler until I know I can implement half pedaling.
I've implemented, and probably will now try to optimize the decay and frequencies as needed. I haven't activated the VSL libraries with SR yet, but hear that they're pretty lacking. I'll have to try myself.
VSL has more reverb and EQ effects. I just use plugins if I want that (VSL is more convenient)
I don't have this at all. This is the first thing to fall out because of the 80/20 rule. I might do this with my own piano since I have unlimited time with it.
Finally, I'm thinking about releasing my workflow (youtube video) and my templates (on github) next year. I'm really standing on the shoulders of others. I've just keep adding more and more features and now have something that might be considered the Uber-template for a DIY Piano Sample library.