The Prompt Snippet Library: How to Stop Rewriting the Same Prompts

There is a prompt that every jewelry designer rewrites at least a dozen times a week. Something like "18k yellow gold, high polish finish, pavé diamond band, photorealistic render, soft studio lighting." It's not complicated. It's just tedious to type every single time, and it's slightly different every time you type it — which means slightly different results.

The Snippet Manager exists to solve this. Save a prompt once, find it instantly, share it with your team, and use it consistently across every project that needs it.

What a snippet actually is

A snippet is a saved prompt with a name, a description, and a set of keywords that trigger it in autocomplete. It can be associated with a specific AI action — so a snippet you built for the Variation tool is tagged to that action and surfaces in context. It can be set to private (just you) or shared across your entire organization.

Snippets can also include image URLs. If you have a reference image that you routinely use as a style guide — say, a photo of a client's existing collection, or a lighting reference you always come back to — you can embed that URL in the snippet. When the snippet is used in a generation, the reference image is automatically forwarded to the AI as visual context. Your style reference and your text prompt travel together.

How autocomplete works in the prompt bar

Snippets act as full, pre-saved prompts that you can insert directly, saving you from typing the entire prompt from scratch. To trigger the feature, simply type a colon (:) followed by your keywords to quickly search through your saved prompts and select the one you need. You also have the flexibility to add extra context or additional text before and/or after the snippet.

Selecting a suggestion fills the prompt bar with the full snippet text. If the snippet includes reference images, they're queued silently in the background. You can then edit the filled prompt to adjust it for the current design before submitting.

Snippet visibility and team sharing

Private snippets are visible only to the person who created them. Organization snippets are visible to every member of the team.

The practical implication is that your best prompts — the ones that produce the most consistent, on-brand results — can become organizational assets rather than individual knowledge. A senior designer builds the snippet for the house's signature metal finish description. Everyone on the team uses it. Outputs stay consistent without anyone having to remember the exact wording.

This also means new team members get up to speed faster. Instead of spending their first weeks learning which prompt phrasings work and which don't, they inherit the accumulated knowledge of everyone who came before them.

Project-Level Instructions: the snippet that runs itself

There's a related feature worth knowing about alongside snippets: Project-Level Instructions.

Where a snippet is something you actively choose to insert into a prompt, a Project-Level Instruction is something that runs automatically on every generation within a project. You set it once per project, and it applies without any action needed from you or your team.

This is the right tool for brand-consistency constraints. If a project is for a client whose collection is always white gold, always with a specific stone cut, always with a certain level of surface texture — you write that as a Project-Level Instruction and every generation in that project inherits it. No one has to remember to include it. No one can accidentally forget it.

Project-Level Instructions can also reference a specific collection by name, which tells the AI to treat that collection's existing designs as a style baseline when generating new ones. This is how you lock an entire project to a client's established aesthetic without manually referencing it in every single prompt.

Voice prompts: when typing is the bottleneck

One more feature that connects naturally here: Voice Prompt Input.

If you find yourself dictating the same descriptions repeatedly — especially for detailed technical prompts that are long to type — you can tap the microphone button in the prompt bar and speak your prompt instead. The recording is transcribed and placed into the prompt field automatically, up to a two-minute maximum.

For designers who work quickly and find typing a friction point, this pairs well with snippets. Use voice for the unique, design-specific parts of a prompt, and let autocomplete fill in the standard technical descriptions from your snippet library.

Building a snippet library that actually gets used

The snippets that see the most use in practice are the specific, technical ones — not the generic ones. A snippet called "photorealistic render" is too broad to be useful. A snippet called "pavé round brilliant half-eternity band, 18k white gold, 1.2mm stone diameter, bright-cut settings, polished finish" is something you'll reach for every time that design type comes up.

Start by auditing your last twenty or thirty generation prompts. Find the phrases that appear repeatedly. Those are your first snippets. Give them sharp keywords that match exactly what you'd type when you want them. Keep descriptions concise so the autocomplete list is scannable.

For organization snippets, have one person own the canonical versions — so there's a single source of truth for how your team describes your standard finishes, metal types, and stone styles. Review and update them when your standard outputs drift, or when you find a new phrasing that reliably produces better results.

The snippet library compounds over time. The more your team uses it and refines it, the more consistent your output becomes — and the less cognitive overhead each generation requires.