Project MAESTRO
Posted: Jun 18, 2017 10:59 pm
Hey all,
The interest in Sonuscore's "The Orchestra" and Spitfire's "Bernard Hermann Toolkit" reminded me of an idea that has been tumbling around in the back of my mind for a few months now.
I'm posting about it here on TSB first and then maybe on VIC to gauge interest.
I will probably put a few months of work into Ask MAESTRO in my free time just to kick it off, and then start a Patreon where the more donations I get per month, the more time I can devote to expanding it. One thing I’m determined about is that Ask MAESTRO will be a free website, never a subscription service.
Ask MAESTRO will be a Machine-Assisted Educational Searchable Typology of Real Orchestrations.
So let me explain…
What is the need for Ask MAESTRO?
The idea behind Ask MAESTRO is this: in the old days you had to go to a concert to learn orchestration, or better yet, work with a session orchestra every day for decades like Josef Haydn or John Williams. Now you can sit at home and study recordings and scores.
But the next leap (from paper to computer) has yet to be made. For example, ask yourself “What does it sound like when a muted trumpet and an English horn play in unison at mezzoforte?” Unless you have an encyclopedic memory, no listenable example might leap to mind. If you want to hear exactly what that sounds like you would have to search page by page through all your scores, or just trust your gut and hope it works on the recording stage. Existing texts about orchestration stick to very broad principles and rules of thumb, and can be unhelpful in these specific/rare situations.
To be blunt, many composers today (myself included!) have inferior music educations & less experience with orchestras than the film composers of previous generations. Other than informal apprenticeships and mentorships, there’s no systematic passing down of knowledge from one gen to the next. That is one reason why lots of music today is less sophisticated and variable than scores of the 70s-90s.
So if you have an orchestration question, you can't exactly dial up Jerry Goldsmith or John Williams. But you can Ask MAESTRO.
How does Ask MAESTRO take the next transformative step in Orchestration education?
MAESTRO relies on what computers are good at, indexing and searching databases. The idea is that any piece of music can be broken down into a succession of orchestration “moments,” in which a particular combo of instruments are playing. Any “moment” can be transcribed using a unique orchestration-oriented shorthand. Then, the moment can be stored with hundreds of others in a database. A website front-end interface can then be used to query the database. Complex queries are enabled by AND, NOT, and OR commands.
For example you might Ask MAESTRO: “I want to hear examples of trumpet and clarinet playing melodies in octaves,” and MAESTRO would spit out six or seven examples from real film scores. MAESTRO would also serve up timestamped Youtube links or some other reference to let you listen to these excerpts.
MAESTRO would be an open-ended inspiration toolkit. For example if you know you want your melody to have violins and oboes in unison at a particular register, you enter that information and MAESTRO can give you real examples of how you might orchestrate the accompaniment and bass line. Or if you're unsure of an instrument combination, you can enter it into MAESTRO and see if it's been used or if it's a (perhaps inadvisable) novelty.
What material will MAESTRO include in its database?
I am going to start Ask MAESTRO by feeding it a front-to-back encoding of the entire score of John Williams’ score Harry Potter And The Sorcerer’s Stone. I like this score because it’s very colorful and has a good balance of wind, brass and string writing. After that I will probably encode E.T. The Extra-Terrestrial or Star Wars V The Empire Strikes Back, or I might take a poll to see what scores people want.
What orchestration information will MAESTRO encode?
I am still developing the shorthand language that MAESTRO will use, but I have decided there are four dimensions that are important to encode. Remember, MAESTRO is about encoding orchestration combinations, not compositional techniques or harmonic language or anything else.
1. Break moments into chunks Any orchestration moment will be composed of “chunks.” A chunk is a unified, single musical idea performed by some combination of instruments - such as a melody, a rhythmic accompaniment, a set of pad-like chords, a scalar decorative run, etc. The chunk types will probably be Theme, Rhythm, Harmony, Decor, and Bass. A moment might contain more than one chunk of each type (for example if there is a melody and a counter-line, these would be considered chunks Theme1 and Theme2). The main reason to start with chunks is because you orchestrate differently when you want an idea to be in the foreground, the background, or a decorative element.
2. Assign instruments to chunks Any chunk would be “populated” by a set of instruments. Each instrument will have a unique three letter codon (such as HRP, EHN, or VN1) and there will be distinct codons for winds/brass playing a1, a2, or triadically. Also there will be distinct codons for arco and pizzicato for the string sections. I’m not planning to go further in detail than that.
3. Assign register to instruments I have broken down the grand staff into 10 overlapping “registers” which are each an octave wide (C1-to-C2, G1-to-G2, C2-to-C3, and so on). This is important because instruments have different sounds in different parts of their range. Every instrument variable in a given chunk is assigned a number that represents where in the staff it plays for most or all of that orchestration-moment. It won’t be an exact science but it will be sufficient to distinguish, for example, the three registers of the flute. The registry number will also allow the computer to “understand and remember” the difference between instruments playing in unison or near-unison (like thirds), versus instruments playing in octaves.
4. Assign dynamics to instruments. Every instrument will have a dynamic assigned to it, whatever it plays for most or all of that orchestration-moment. Again not an exact science but it’s important because instruments have very different balances at different dynamics as exemplified by the old Rimsky-Korsakov rule about doubling the horns at higher dynamics.
How will the front end work?
Still brainstorming about this as well, but I picture a front-end website where you can click on pictures of instruments to select and deselect them. MAESTRO then searches for moments in its database that match up to what you asked to hear. Using dropdown menus you can assign instruments to particular ranges (e.g. I want to hear a high flute not low flute; I want to hear the flute in octave above english horn, etc). Once the orchestration-moment is encoded and in the database, it’s possible to do very powerful, variable, and specific searches depending on what the user wants to know.
Why John Williams?
JW is acknowledged by pretty much everyone as the living master of orchestration for film music. I like the idea of keeping the database specific to his works, but I might consider adding other great film composers if it’s a popular idea and I can find scores.
The interest in Sonuscore's "The Orchestra" and Spitfire's "Bernard Hermann Toolkit" reminded me of an idea that has been tumbling around in the back of my mind for a few months now.
I'm posting about it here on TSB first and then maybe on VIC to gauge interest.
I will probably put a few months of work into Ask MAESTRO in my free time just to kick it off, and then start a Patreon where the more donations I get per month, the more time I can devote to expanding it. One thing I’m determined about is that Ask MAESTRO will be a free website, never a subscription service.
Ask MAESTRO will be a Machine-Assisted Educational Searchable Typology of Real Orchestrations.
So let me explain…
What is the need for Ask MAESTRO?
The idea behind Ask MAESTRO is this: in the old days you had to go to a concert to learn orchestration, or better yet, work with a session orchestra every day for decades like Josef Haydn or John Williams. Now you can sit at home and study recordings and scores.
But the next leap (from paper to computer) has yet to be made. For example, ask yourself “What does it sound like when a muted trumpet and an English horn play in unison at mezzoforte?” Unless you have an encyclopedic memory, no listenable example might leap to mind. If you want to hear exactly what that sounds like you would have to search page by page through all your scores, or just trust your gut and hope it works on the recording stage. Existing texts about orchestration stick to very broad principles and rules of thumb, and can be unhelpful in these specific/rare situations.
To be blunt, many composers today (myself included!) have inferior music educations & less experience with orchestras than the film composers of previous generations. Other than informal apprenticeships and mentorships, there’s no systematic passing down of knowledge from one gen to the next. That is one reason why lots of music today is less sophisticated and variable than scores of the 70s-90s.
So if you have an orchestration question, you can't exactly dial up Jerry Goldsmith or John Williams. But you can Ask MAESTRO.
How does Ask MAESTRO take the next transformative step in Orchestration education?
MAESTRO relies on what computers are good at, indexing and searching databases. The idea is that any piece of music can be broken down into a succession of orchestration “moments,” in which a particular combo of instruments are playing. Any “moment” can be transcribed using a unique orchestration-oriented shorthand. Then, the moment can be stored with hundreds of others in a database. A website front-end interface can then be used to query the database. Complex queries are enabled by AND, NOT, and OR commands.
For example you might Ask MAESTRO: “I want to hear examples of trumpet and clarinet playing melodies in octaves,” and MAESTRO would spit out six or seven examples from real film scores. MAESTRO would also serve up timestamped Youtube links or some other reference to let you listen to these excerpts.
MAESTRO would be an open-ended inspiration toolkit. For example if you know you want your melody to have violins and oboes in unison at a particular register, you enter that information and MAESTRO can give you real examples of how you might orchestrate the accompaniment and bass line. Or if you're unsure of an instrument combination, you can enter it into MAESTRO and see if it's been used or if it's a (perhaps inadvisable) novelty.
What material will MAESTRO include in its database?
I am going to start Ask MAESTRO by feeding it a front-to-back encoding of the entire score of John Williams’ score Harry Potter And The Sorcerer’s Stone. I like this score because it’s very colorful and has a good balance of wind, brass and string writing. After that I will probably encode E.T. The Extra-Terrestrial or Star Wars V The Empire Strikes Back, or I might take a poll to see what scores people want.
What orchestration information will MAESTRO encode?
I am still developing the shorthand language that MAESTRO will use, but I have decided there are four dimensions that are important to encode. Remember, MAESTRO is about encoding orchestration combinations, not compositional techniques or harmonic language or anything else.
1. Break moments into chunks Any orchestration moment will be composed of “chunks.” A chunk is a unified, single musical idea performed by some combination of instruments - such as a melody, a rhythmic accompaniment, a set of pad-like chords, a scalar decorative run, etc. The chunk types will probably be Theme, Rhythm, Harmony, Decor, and Bass. A moment might contain more than one chunk of each type (for example if there is a melody and a counter-line, these would be considered chunks Theme1 and Theme2). The main reason to start with chunks is because you orchestrate differently when you want an idea to be in the foreground, the background, or a decorative element.
2. Assign instruments to chunks Any chunk would be “populated” by a set of instruments. Each instrument will have a unique three letter codon (such as HRP, EHN, or VN1) and there will be distinct codons for winds/brass playing a1, a2, or triadically. Also there will be distinct codons for arco and pizzicato for the string sections. I’m not planning to go further in detail than that.
3. Assign register to instruments I have broken down the grand staff into 10 overlapping “registers” which are each an octave wide (C1-to-C2, G1-to-G2, C2-to-C3, and so on). This is important because instruments have different sounds in different parts of their range. Every instrument variable in a given chunk is assigned a number that represents where in the staff it plays for most or all of that orchestration-moment. It won’t be an exact science but it will be sufficient to distinguish, for example, the three registers of the flute. The registry number will also allow the computer to “understand and remember” the difference between instruments playing in unison or near-unison (like thirds), versus instruments playing in octaves.
4. Assign dynamics to instruments. Every instrument will have a dynamic assigned to it, whatever it plays for most or all of that orchestration-moment. Again not an exact science but it’s important because instruments have very different balances at different dynamics as exemplified by the old Rimsky-Korsakov rule about doubling the horns at higher dynamics.
How will the front end work?
Still brainstorming about this as well, but I picture a front-end website where you can click on pictures of instruments to select and deselect them. MAESTRO then searches for moments in its database that match up to what you asked to hear. Using dropdown menus you can assign instruments to particular ranges (e.g. I want to hear a high flute not low flute; I want to hear the flute in octave above english horn, etc). Once the orchestration-moment is encoded and in the database, it’s possible to do very powerful, variable, and specific searches depending on what the user wants to know.
Why John Williams?
JW is acknowledged by pretty much everyone as the living master of orchestration for film music. I like the idea of keeping the database specific to his works, but I might consider adding other great film composers if it’s a popular idea and I can find scores.
) that John Williams emphasizes the chalumeau and clarion registers and writes relatively little for the middle register. Having register info for his entire clarinet use across a score lets me test that hypothesis, and also look at whatever middle-register excerpts there are and see if I can learn the reason why he used middle clarinet at that point. And so on.