Presets & AI

Palora's presets are an open, text-based format. Every preset is a JSON file (.palora), and a bank is a single .palorapack file holding all its presets.
This open format enabled us to write an AI authoring brief for making single presets, or complete preset packs for Palora. Paste the brief into your AI, describe the sound you want, and import the preset-JSON directly into Palora.

Making presets with AI

Teach the assistant the engine once, then ask for sounds.

1. Get the brief. Download the AI brief (authoring-for-ai.md) — a self-contained guide that teaches any capable assistant Palora's engine and preset format.
2. Paste it into your assistant's context — Claude, ChatGPT, Gemini, whichever you already use.
3. Ask for what you want (the recipe below), then paste the result back into Palora. Importing works during the trial and with a license, not after the trial has ended.

The Claude skill

If you work in Claude, there is a second route. Install the Palora preset skill and Claude picks up the same knowledge on its own, loading only the part it needs for what you asked — no long paste at the top of every conversation, and it stays available in new chats without you setting it up again.

The skill carries the same brief, so the presets it writes are the same format.

Download the Palora preset skill (ZIP)

Unzip it into ~/.claude/skills/palora/ for Claude Code, or add the folder through the Skills UI in Claude Desktop. The README.md inside covers both, and keeping the folder structure intact is what lets Claude load one reference at a time.

Everything below applies to both routes.

How to ask

What you get back is mostly decided by how you ask for it. The three parts below are one thread — the habits, an example that uses them, and what lands when you send it.

Five things that get you good presets

Real banks — not sixteen renamed clones of one patch — come down to how you ask.

1. Name the sub-roles, or ask the assistant to propose them first. "16 plucks" invites sixteen variations of one preset. Instead: "16 plucks: 3 piano, 3 glass, 3 organic, 3 chord, 2 bell, 2 experimental." If you don't know the breakdown, say "propose a 16-pluck role list first, and wait for my OK before generating."

2. Give a vibe, a genre, or a reference. "lo-fi, dusty, late-night," or "the kind of pluck that carries a downtempo track," anchors character far better than adjectives alone.

3. Chunk big banks. Assistants template and drift over one giant output, so ask for one category (or roughly 8–12 presets) per paste. You get sharper results and can audition each batch before the next.

4. Make it show its plan before the JSON. *"List each preset's musical role

   • macro idea + motion idea first; I'll approve, then you emit the bank."* That catches clones before the output is spent generating them.

5. Iterate by ear. The first draft won't be perfect. Load it, listen, then come back with targeted notes — "#3 is too dark, #7's macro does nothing" — and ask for a revised bank. Or even better: use the presets as a boilerplate for your own imagination.

A good prompt, end to end

"Make me a 10-preset Bass bank, deep-house, warm and round. Sub-roles: 3 classic sub-basses, 3 plucky stabs, 2 reese-ish wide basses, 2 distorted/driven. List each preset's role + macro + motion idea first and wait for my OK, then emit one .palorapack bank."

Ask for one sound and the same habits apply, minus the roles:

"Make me one plucky deep-house bass — short, round, a little driven. Put the filter's brightness on Macro 1."

What you get back

• A preset, or a bank — a single .palora file when you asked for one sound, a .palorapack when you asked for a pack. Either way it is one paste into Palora, and a bank lands as a navigable pack of its own.
• A short rationale describing the design intent — for a bank, how the presets differ from one another.
• A first draft. Treat the first emit as a starting point to refine by ear, not a finished product.