I try to figure out PYT of Michael Jackson to look at the exact Greg Phillinganes Rhodes chords and the voicing but none of ChatGPT, Gemini, Grok, hook theory and YouTube piano tutorials have the same insight for the answer!?
It seems difficult to trust someone if you don't have very good ears...

 

They're all still very bad a that, yeah. I'm not sure why but I'm dead sure I wouldn't trust them for anything music theory related... maybe music history.

 

a LLM probably thinks Beethoven is a Saint Bernard, then will tell you're special for asking.

 

Can you try splitting stems, then loading into Melodyne?

Try splitting stems, then feed to AI?

 

Perhaps AI LLMs struggle
For theory voicing in a song
Because...
It is not actually down to
Dry mathematical formulae, and
Even though they cannibalize
The physical notes of musicians
They can never digest our souls.

 

Last edited: Apr 8, 2026 at 1:43 PM

Well, I'll be a bit off topic, but I don't trust much tools like melodyne and similar, they get the basic line, but nuances are hard to get, and I don't ask help to AI for that matter.

However, i must say I was astonished by a software from Aurally Sound - Prism.
I thought was yet another trascribe tool comparable to all the others, but the results I got with it were jaw-dropping.
I played a pretty complex piano piece, and it mostly got it all right, beside some sustain pedal passage that could shuffle notes.
If I can i'll post original and the reconstruction.

This tool is only for solo instruments.

 

Exceptionally, here is the AI answer:

Spoiler: Perplexity KI

What’s tripping you up is that PYT exists in more than one musical “version”: the released Thriller track, the earlier Michael/Greg demo, and then transcription guesses that simplify the keyboard part into generic chords. The demo is credited to Michael Jackson and Greg Phillinganes, while the released song was reworked by Quincy Jones and James Ingram, so the exact Rhodes voicings people share online often don’t match each other.

Why the answers disagree
Most tutorial sites reduce the part to the harmonic shell, not the actual voicing. For example, one common chord transcription gives the song as B7, Em7, G, A, which may describe the progression but not the exact Rhodes shape or inversion. Another analysis notes the song is harmonically more complex than average and centers around B Dorian / B minor regions, which also helps explain why a simple four-chord answer can feel incomplete.

What the Rhodes part likely is
The Rhodes comping in PYT is best understood as stacked voicings and voice-leading, not block chords copied from a chart. In practice, that usually means:

left hand holding root or rootless support,
right hand using 3rds, 7ths, 9ths, and sometimes sus tones,
smooth movement between shapes instead of big jumps.
So when people say “it’s B7 to Em7,” they may be naming the harmony correctly while missing the actual keyboard color that makes Greg’s part sound right.

How to get closer to the real voicing

A reliable way to reconstruct it is:
Identify the bass note at each hit.
Listen for the guide tones first, especially 3rd and 7th.
Add color tones like 9ths or 11ths only after the shell works.
Match the comping rhythm before worrying about note-for-note density.
That approach usually gets closer to the original than copying a piano tutorial with simplified chord names. The reason is that Rhodes parts in records like this are often arranged as voiced textures, not “lead-sheet chords”.

Practical takeaway
If your goal is the exact Greg Phillinganes feel, you probably need to treat the part as a transcription job, not a chord-chart job. Start from the progression B7 / Em7 / G / A as a harmonic skeleton, then build a voicing-by-voicing ear transcription over the recording, because the missing detail is in the inversion, spacing, and rhythmic placement rather than just the chord symbols.

A useful next step is to isolate one short phrase of the song and transcribe only the Rhodes attack points; that will reveal the actual voicing much faster than trying to decode the whole track at once.

 

That's why most automatic tools fail in giving the full transcription (but Prism was near).

That BTW are roughly the steps I do to transcribe (not always sorted this way):
1 - get the bass, so just from the I and V you'll get the base scale (really!), no matter what type is;
2 - guess if major or minor;
3 - guess if a modal scale (not very important, you'll get it in step 4 and 5);
4 - guess the melody (usually just monophonically);
5 - guess basic triad chords playing along the bass (like Am, C, etc.);
6 - guess the chord inversions if any (in most cases they are);
7 - guess if there are 7th, 9th, 11th, sus, and so on;
8 - arrange with the right voicing (arpeggio, etc.) and add harmony to the melody if any.

For all these steps, unless you want to hit the keyboard hoping that chance is with you, you need at least some basics in music theory aka harmony theory.

Edit: I forgot: there are times that songs modulate to other scales, that complicates a little, but not that much.

 

AI gives better responses with specific, detailed questions. If you can break down what you're hearing and convey what you think is being played in words it will get you closer, just keep hammering it with questions, telling it that it's wrong and it will keep analysing. I'm talking about a chat AI but in my experience they're very good at this kind of interaction. For example, AI will be able to give you a great deal of insight into Greg Phillinganes style, usual choices. It can't analyse audio but it doesn't need to, it already 'knows' the song structure of PYT and the exact chords Greg plays - eg., (the following is Copilot AI):

Here’s a clean, accurate, musician‑level breakdown of Greg Phillinganes’ style — the kind of insight an AI can give, because this is pattern‑based, not audio‑based.

I’ll structure it the way a session player would explain another session player.

Spoiler

Greg Phillinganes: Core Style Traits
1. Ultra‑clean, ultra‑precise voicings
Greg’s Rhodes and synth parts are surgically tidy.
He avoids muddy clusters and prefers:

• 3‑note shells (rootless when the bass is covered)
• tight 4‑note voicings with a clear top voice
• spread voicings with 10ths or 6ths in the left hand
• upper‑structure triads for colour (especially over dominant chords)

He plays like someone who knows the exact harmonic function of every note.

2. Pop‑funk economy
Greg never overplays.
He’s the opposite of a gospel‑chops keyboardist.

His parts are:

• short
• rhythmic
• intentional
• supportive

He plays hooks, not “keyboard parts”.

Listen to any MJ track he’s on — the Rhodes is basically a rhythmic instrument.

3. The “Phillinganes snap” on Rhodes
He has a very recognisable Rhodes attack:

• slightly harder velocity
• bright tine
• tight envelope
• almost clav‑like articulation

He often accents the off‑beats with a little extra bite.

This is why his Rhodes parts cut through dense mixes without being loud.

4. Syncopation that locks with the drums
Greg plays with the drummer, not over them.

Typical traits:

• anticipations on the “e” or “a”
• ghosted left‑hand notes
• right‑hand stabs that mirror the guitar
• rhythmic displacement (he loves pushing a chord 1/16th early)

On PYT specifically, the Rhodes is basically part of the percussion section.

5. He outlines harmony with minimal notes
Greg often implies full chords with:

• 3‑note voicings
• dyads
• octave + colour tone
• triads over alternate bass notes

He trusts the bass player and the arrangement.

This is why his parts feel so clean.

6. Signature harmonic tendencies
He gravitates toward:

• major 9
• minor 11
• dominant 13
• sus2 / sus4
• add9
• quartal voicings in passing
• chromatic approach chords (very subtle)

He rarely plays plain triads unless the arrangement demands it.

7. Synth parts that are “keyboardist clean,” not “sound designer messy”
Greg’s synth lines are:

• monophonic
• precise
• hook‑driven
• rhythmically tight
• harmonically simple but placed perfectly

He doesn’t noodle.
He plays statements.

How this applies to PYT
Without analysing the audio directly, we can still say:

• the Rhodes part is tight, syncopated, and minimal
• the voicings are clean, upper‑mid register, no mud
• he uses add9 and 6/9 colours
• the left hand is light, mostly outlining roots or 10ths
• the right hand plays stabs, not pads
• the part is locked to the drums, not floating

This is classic Greg.

⭐ If you want, I can go deeper
I can break down:

• typical Greg Phillinganes voicing shapes
• his left‑hand habits
• his right‑hand rhythmic vocabulary
• how he approaches MJ‑era Rhodes comping
• what he tends to do on dominant chords
• how he voices minor chords in pop‑funk

Or I can give you a Greg‑style voicing dictionary you can use to reverse‑engineer PYT.

Just tell me which direction you want to go.

 

But that requires too a knowledge of (basic or advanced - it depends) music theory.
I guess many want the supper ready instead.

 

Last edited: Apr 8, 2026 at 12:21 PM

Try to get the original studio stems. Maybe they are even on YT or somewhere. Use surgical EQs for further analysis of the track if needed.

 

Why don't you ask Greg?