Heinz Moser

When composing AI music, a key aspect is prompting—giving software like Suno instructions on how the AI should play the piece, which instruments to use, how the singer should perform, and so on.

People often obtain such stylistic elements by asking an AI assistant to fill in the field with stylistic elements in Suno, because the chatbot has a vast number of prompts at its disposal thanks to its generative LLM model. There are now also a number of resources (books, videos, blogs) that describe how to prompt Suno. A search on “ChatGPT” turned up a whole range of books and articles that can be consulted here (see the end of this post).

However, the chatbot immediately pointed out that there are more useful things to do than working through lists from the prompt that you might eventually adopt. After all, you’re probably not just looking for a list of Suno tags, but want to learn how to give Suno instructions in general that improve the song and tailor it to your desired style. According to ChatGPT, when using the Celtic rock example, you should look for guidance on the following questions:

– Which Celtic rock subgenre fits best?

– What instrumentation could be chosen?

– What drum sound is best?

– Which guitars should be used?

 – What bass playing techniques are effective?

– Which Celtic instruments should be included?

– What should the vocal characteristics be like?

According to ChatGPT, there is now some interesting empirical evidence regarding such questions. For example, experienced Suno users recommend not simply cramming in as many keywords as possible, but rather structuring the prompt into a few musically meaningful layers.

ChatGPT therefore recommends not simply following instructions, but rather using the AI as a kind of advisor. For example, ChatGPT suggested to me: “For your Shadow Light project, I would take a different approach: I could create a ‘Suno Style Prompting Guide for the Shadow Light Project’ for you—essentially your own personal reference guide.”

Chat GPT's suggestion: "If I were you, I wouldn't stop at 'Suno Prompting.' We could use these sources to develop a kind of 'Shadow Light Prompting System.'"

The post clearly illustrates how ChatGPT envisions this: ChatGPT begins by distinguishing between different musical styles within the Celtic rock genre (a similar approach would, of course, also be possible for German Schlager, hard rock, hip-hop, etc.).

 When it comes to Celtic rock, ChatGPT distinguishes between:

"Celtic Rock

→ 8–10 musical parameters

Political Celtic Ballad

→ custom parameters

Irish House

→ custom parameters

Celtic House

→ custom parameters

Celtic Rock Anthem

→ custom parameters

and for each style, respectively:

Genre → Groove → Tempo → Instruments → Guitars → Celtic Instruments → Bass → Drums → Vocals → Chorus → Harmony → Song Structure → Production → Exclusions.»

This leads to very practical instructions for users. It would have “a major advantage: In the future, you could simply tell me, ‘Make me a Shadow Light-style prompt, but this time darker, 105 BPM, more uilleann pipes, and a more aggressive chorus’; and we could systematically build on your established sound instead of coming up with a completely new Suno prompt every time. Especially for your Celtic rock songs, I think this approach is far more promising than any off-the-shelf prompt book.”

In its “Prompting System,” the bot does not view it as a book of pre-made prompts, but rather as a system of building blocks that allow users to specifically control their own Shadow-Light-Sound.

For the Shadow Light Project, ChatGPT plans to build such a system on three levels:

"1. The Shadow-Light-Sound Identity"

First of all, there’s the “fingerprint,” which ideally remains consistent across several songs:

Basic Character

• modern, emotional, organic

• Celtic/folk roots

• Rock as the musical backbone

• an authentic live-band aesthetic

• not just any “fantasy-Celtic” atmosphere

• more European/Irish than Hollywood-style Celtic

Instrumental DNA

• acoustic guitar

• electric rhythm guitars

• distinctive electric bass

• organic drums

• Fiddle

• Uilleann pipes / tin whistle: specifically, not all the time

• occasionally, bodhrán

• acoustic percussion

Vocal DNA

• male voice

• more like a baritone/tenor

• slightly rough and full of character

• a believable, understated interpretation

• Verses that are more narrative in nature

• A bigger, more emotional chorus

• occasional polyphonic pub/hall chorus

2. The Shadow-Light Style Matrix

Each song is then assigned scores for these categories:

Parameter Possible Values

Genre Celtic Rock / Folk Rock / Irish Rock / Celtic House

Tempo slow / mid-tempo / fast

Groove straight / driving / shuffle / dance

Energy 1–10

Darkness 1–10

Emotionality 1–10

Instrumentation Folk ↔ Rock

Celtic influence subtle ↔ dominant

Guitars acoustic ↔ heavy

Drums folk-inspired ↔ stadium-style

Bass subdued ↔ driving

Vocals intimate ↔ hymnal

Chorus solo ↔ backing vocals

Production raw ↔ polished

Live Feel Studio ↔ Live

Dynamics steady ↔ sharp increase

This allows us, for example, to define a song as:

Celtic Rock / 106 BPM / Energy 8 / Darkness 6 / Emotionality 9 / Rock 7 / Celtic 5 / Vocals 8 / Chorus 10 / Live 8

To me, that's much more meaningful than ten random adjectives.

3. The actual Suno prompt

From that, I would then create a concise style prompt, such as

Modern Celtic rock, 106 BPM, driving live drums and melodic electric bass, acoustic guitar foundations with gritty, distorted rhythm guitars, expressive fiddle and occasional uilleann pipes, a dark yet uplifting minor-key atmosphere, male baritone lead vocals with a slightly weathered quality, restrained, intimate verses building into a massive, anthemic chorus with male gang vocals, organic live-band energy, dynamic arrangement, emotionally powerful yet never glossy or overproduced.

Even more important is this: song structure. Suno doesn't just respond to the overall style. We should therefore have a second layer for the song's progression.

[Intro]

• acoustic guitar + fiddle

• restrained

[Verse 1]

• intimate male vocal

• sparse instrumentation

[Pre-Chorus]

• the bass becomes more prominent

• the drums gradually build

[Chorus]

• full band

• distorted guitars

• backing vocals

• anthemic

[Verse 2]

• stronger drums

• additional fiddle

[Bridge]

• drop-down

• acoustic / vocal focus

[Final Chorus]

• maximum energy

• layered vocals

• full Celtic instrumentation

This allows us to direct a song from a dramatic perspective, rather than simply telling Suno, “Make a Celtic rock song.” I would specifically include “negative steering” in this system: not just telling Suno what to do, but also—when dealing with problematic styles—what to avoid.

For example:

no EDM drops, no glossy pop production, no cinematic Hollywood choir, no fantasy soundtrack vibe, no excessive synths, no overly clean vocals, no generic folk-rock

Finally, we should create several "Shadow-Light" style profiles

I would start by developing these six presets:

SL-01 — Modern Celtic Rock

The actual base sound.

SL-02 — Celtic Political Ballad

For songs like your political and social ballads.

SL-03 — Celtic Rock Anthem

Great chorus, vocal harmonies, festival/live vibe.

SL-04 — Dark Celtic Rock

Darker, rougher, at times almost post-punk.

SL-05 — Irish/Celtic House

The style of “Oh My Johnny”—traditional elements with a dance groove.

SL-06 — Celtic Storytelling

Less rock, more of a narrative folk vibe.

Next comes the most interesting part of the consultation with ChatGPT: We could analyze each of your already successful Suno songs in reverse. So not: “What would be a good Celtic rock prompt?”

but rather: “Why does this particular song work?”

To do that, I would, for example, break down one of your hit songs something like this:

Genre: Celtic Rock

Speed: 108

Drums: tight live rock

Bass: melodic/driving

Acoustic guitar: prominent

Electric guitar: crunchy, not metal

Fiddle: accents

Vocals: male, weathered, conversational

Chorus: gang vocals

Dynamics: restrained → explosive

Production: organic, broad, slightly raw

Then the true Shadow-Light DNA emerges from your own successful songs. In my opinion, that would be much better than recreating a generic Suno manual.

Conclusion:

In this first part, I used ChatGPT as an example to highlight—and support with quotes—how this AI envisions supporting users. What is presented here is a strategic work package. Ultimately, it is intended to help users apply successful concepts right from the start of their songwriting process. Of particular interest is the aspect of conducting a backward analysis of songs that have already been released—in order to substantiate their success or failure based on various criteria.

However, the yardstick by which songs are measured is their commercial success. Yet landing in the Top 10 of the charts solely on the strength of a cleverly crafted prompt arrangement seems highly unlikely to us in a situation where 97 percent of all AI-generated songs go unnoticed. Furthermore, even if it is the creators’ goal to appeal to as many people as possible with their songs in order to monetize their work through the number of subscribers they gain, that is only one aspect of a song’s quality. For example, someone who focuses on a relatively niche genre with a small number of interested fans—and isn’t even aiming for a large number of subscribers—might set entirely different quality criteria.

In the second part of this report, I will discuss the criteria that other AI systems use when supporting songwriters. It will become clear that AI assistants approach supporting creators using a wide variety of methods. ChatGPT’s modular approach is just one example. Some systems use different criteria—which, in turn, influences the results. In any case, these bots are not simply neutral and interchangeable tools where it doesn’t matter which one you choose. But more on that in Part 2 of this blog post…

The quotes from ChatGPT are from my queries on August 18, 2026

ChatGPT recommends the following literature, among others

🥇 1. Blake Crosley – “Suno Guide: Tags, Meta Tags & Prompts”
This is currently my favorite free comprehensive guide.
Suno Guide: Tags, Meta Tags & Prompts (V5.5)

🥈 2. HookGenius – “The Suno Mastery Guide”
This is interesting because it’s specifically offered as a PDF and has since been updated to version 5.5.