Artist Research Loop
Research an artist deeply once, store them in your Artist Library, then channel that voice on every future song — so the study compounds instead of repeating.
Anyone who writes in real artists' voices and is tired of shallow, samey AI output — you want each artist you channel grounded in genuine research that is saved, reusable, and wired into the sub-genres they actually sit in.
This loop solves a quiet waste at the heart of AI songwriting: every time you decide to write "in the style of" some artist, the agent re-guesses who they are from scratch — and guesses shallow. The result is the same beige imitation every time. This loop does the research properly once, stores it as a real record in your Artist Library, and then RETRIEVES that record on every future song you write in that voice. The study compounds; the slop dies.
When you name an artist (or your agent picks a voice to channel for a song), it studies them for real: who they are, the arc of their career, why they matter, and how popular or influential they actually are. It pins down their active years, the broad genres AND the specific sub-genres they live in — linking those sub-genres straight into the site's genre encyclopedia so an artist record and the genre research reinforce each other. It writes one or two paste-ready style prompts that capture their sound for a generator, lists their top songs, and — the part that makes a voice usable instead of generic — separates how they FIT their sub-genres from how they break the mold and do their own thing. That contrast is what gives a write-in-their-voice song an actual edge.
The Artist Library is public and "user-added": a growing, community-built companion to the official genre encyclopedia, kept distinct from it. Your stored artist records live alongside everyone else's, each one tied into the same genre trees, so the library gets richer the more anyone uses this loop. Next time you — or anyone — wants that voice, the research is already there to pull.
This is a research-and-memory loop, not a writing loop. It pairs with the Jarvie Music Writing Loop and the Songwriting Craft Loop: it feeds their "step into an artist's mind" moment and hands their style-prompt step a ready-made, grounded prompt. A hard line throughout: naming a real artist as a study reference is fine and editorial, but the loop never instructs impersonation of a living artist's identity, and the final style prompts describe the sound in craft language and adjectives — never by leaning on a living artist's name.
Skills & actions it uses
The concrete, reusable skills this loop calls to actually do the work.
The Artist LibraryThe public, user-added store this loop is built on — it both SAVES a fully researched artist record (description, active years, genres + sub-genres, style prompts, top songs, fit-vs-differ) and RETRIEVES it on every later song so research is never repeated.The genre encyclopediaResearched alongside the artist — the loop studies the specific sub-genres the artist sits in and links the artist record into the encyclopedia's genre trees, so artist research and genre research reinforce each other.
The loop
- 1
Name the voice and check the library first
Start from an artist — one you named, or one your agent chose to channel for this song. Before any new research, look them up in the Artist Library: if a solid record already exists, skip straight to retrieving and using it. Research is only repeated when it's missing or thin.
- Take the artist name (or the agent's chosen voice) as the input
- Search the Artist Library for an existing record before studying anything
- If a strong record exists, jump to the retrieve-and-apply step — don't redo the work
- Only continue into research when there's no record, or the one there is shallow
- 2
Research the artist for real
Drop into study mode and learn who this artist actually is — not a one-line cliché. The goal is a record specific enough that channeling them produces something with a real edge, not the average of everyone in their lane.
- Write a description: their history, the arc of their career, and why they matter
- Gauge their popularity and influence — niche cult figure vs. genre-defining mainstream act changes how you'd write
- Pin down their active years (when they emerged, their peak, whether they're still working)
- Capture the signatures: vocal character, cadence and phrasing, lyrical themes, production palette, the moves that are unmistakably theirs
- 3
Place them in genres AND specific sub-genres
Locate the artist in the genre map at two resolutions — the broad family AND the precise sub-genres — and wire those sub-genres into the site's genre encyclopedia so the artist record and the genre research strengthen each other.
- Tag the broad genre(s) they're known for
- Name the specific sub-genre(s) they actually live in — precision here is what makes the voice usable
- Link those sub-genres into the existing genre encyclopedia tree so the record connects to real genre research
- If a sub-genre they belong to has a thin or empty encyclopedia entry, that's a signal it's worth researching too
- 4
Pin down how they FIT vs. how they DIFFER
This is the step that makes a voice worth channeling. Separate the ways the artist sits squarely inside their sub-genres from the ways they break the mold — because the breaks are usually what make them sound like nobody else.
- How they FIT: the sub-genre conventions they honor — the instrumentation, structures, and tropes that place them clearly in the style
- How they DIFFER: the deviations that are their signature — an unexpected cadence, a genre they fold in, a production choice that's against type
- Write the contrast explicitly; "channel the differ, not just the fit" is what separates an original take from a karaoke imitation
- Keep it as craft observation about the sound — not an instruction to impersonate the person
- 5
Write 1–2 paste-ready style prompts
Distill everything into one or two tight, generator-ready style prompts that capture this artist's sound — built from sub-genre plus craft adjectives, shaped to drop straight into a generator's style box. This is the artifact the writing loops' style-prompt step will reach for.
- Front-load the sub-genre, then name instruments, vocal mode, and one emotional anchor
- Describe the voice in adjectives and craft language — NEVER by naming the living artist in the prompt
- Bake in the "differ" — the signature deviation — so the render sounds like them and not the genre average
- Keep each prompt tight enough to respect a generator's style-box limit (e.g. Suno's ~200 characters)
- Offer one or two variants when the artist has more than one distinct mode worth capturing
- 6
List their top songs as touchstones
Capture a short list of the artist's most defining or essential songs, so the record carries concrete reference points your agent (or you) can study for cadence, structure, and attitude later.
- A handful of their signature or most essential tracks
- Lean toward songs that best demonstrate the FIT-vs-DIFFER you just described
- These are study touchstones, not material to copy — references for craft, not lyrics to lift
- 7
Store the whole record in the Artist Library
Save the complete record — description, active years, genres and sub-genres, fit-vs-differ, style prompts, and top songs — into the public, user-added Artist Library, linked into the genre encyclopedia. This is the step that makes the research permanent and reusable for you and everyone else.
- Persist every field as one artist record, distinct from the official genre encyclopedia
- Keep the sub-genre links live so the record stays connected to the genre trees
- Because the library is public and user-added, your research enriches the shared library, not just your own session
- Once stored, the next request for this voice can skip straight to retrieval
- 8
Retrieve and feed it into a writing loop
Whenever you write a future song in this voice, pull the stored record and hand it to a writing loop — it grounds the "step into an artist's mind" moment and supplies the style-prompt step with a ready, researched prompt instead of a fresh guess.
- On any later song in this voice, retrieve the record instead of re-researching
- Feed it to the Jarvie Music Writing Loop or the Songwriting Craft Loop as the artist-voice input
- It directly powers their style-prompt step with the paste-ready prompt already built here
- Same artist, many songs, one research pass — the compounding payoff of the whole loop
Then run it again for the next song, chapter, or track.
You end with
- A complete, reusable artist record in the public Artist Library: description (history + why they matter + popularity), active years, broad genres, and specific sub-genres linked into the genre encyclopedia
- An explicit "how they fit their sub-genres" vs. "how they do their own thing" breakdown — the contrast that gives a channeled voice a real edge
- One or two paste-ready style prompts that capture the artist's sound in craft language and adjectives, no living-artist name, ready for a generator's style box
- A short list of the artist's top/essential songs as study touchstones
- A retrievable voice you can pull into the Jarvie Music Writing Loop or Songwriting Craft Loop on every future song — research done once, reused forever
Skip this loop if…
- You want the song's lyrics or finished audio — this loop researches and stores a voice; a writing loop and a create loop turn it into a song
- You want the AI to clone or impersonate a specific living artist's identity rather than study them and write an original voice grounded in that research
- You're writing a one-off and will never revisit this voice — the payoff of this loop is the research compounding across many future songs
How it connects
Pairs with
Commonly run just before or after this one in the pipeline.