Design

AI-assisted style exploration and prompt reuse

Use Claude to retrieve and assemble previously successful Midjourney styles, prompts, and parameters stored in Notion, enabling faster visual exploration and consistent client-ready outputs without manually searching across notes and threads.

Why the human is still essential here

The designer still sets the creative direction, selects what looks right, and decides what becomes the final system; AI speeds up recall, structuring, and iteration.

How people use this

Prompt library retrieval for new concepts

Ask an AI assistant to pull 5–10 past Midjourney prompt "recipes" from a Notion database that match a new project's mood and format them for immediate reuse.

Claude / Notion

Consistent style variants for a brand system

Generate multiple image variations using the same saved prompt parameters (seed, stylize, aspect ratios) to keep a client's visual exploration consistent across a campaign board.

Midjourney

Prompt refactoring and parameter tuning

Have AI rewrite an older prompt into a cleaner template with controllable variables (subject, material, lighting, lens, color palette) while preserving the original look.

ChatGPT

Prompt database with tags and fields

Save prompts, parameters, and sample outputs into a Notion database so past styles can be filtered by client, project, or aesthetic tags.

Notion

Natural-language style recall

Ask for the grainy editorial black-and-white look from last summer and have AI pull the closest stored prompt and settings from Notion.

Claude / Notion

Auto-summarize and standardize prompts

Convert messy prompts into a consistent template (subject, lighting, lens, palette, negatives, params) and write the cleaned version back to Notion.

Claude / Notion

Community stories (1)

X

A Designer's AI Stack

In the past few months I've replaced or automated a lot of the work that used to sit around my design process - managing projects, writing briefs, searching for prompts I made a while back, building interactions I could picture but couldn't code.

Here are the tools that actually made that happen and how I use them to keep track & automate everything boring in my process.




Claude + Midjourney


Midjourney has been a cornerstone for a lot of what I've been doing in the past year or so. The speed of style exploration is hard to match, and once I'm locked in a direction the consistency I can get out of it is something clients are actually happy with. Lately I've had a couple of people reach out specifically wanting a visual style/system they could use for their own content.


What changed how I use it was bringing Claude into the process on the prompt side. I learned early to store favourite prompts, profiles, images and styles somewhere, but they'd end up spread across notes, different Notion pages, message threads - everywhere. Now they live in Notion in a structured way, and since Claude already has access to that, I can just ask for a style I built months ago instead of searching for it.


This also works for helping me put together prompts and styles in batches for clients. Instead of handing someone a finished image, I can give them a repeatable structure - a set of parameters and style descriptors that means every image they generate looks on-brand, even if they've never written a Midjourney prompt before. It removes a lot of back and forth and gives clients something useful as an additional deliverable.




A custom Gem agent


Workshop is probably my favourite Framer feature of them all - being able to create complex code components with simple prompts is a huge deal and it handles most of what I need day to day.


But I've always loved WebGL shaders, advanced scroll interactions, infinite galleries, or anything with more complex logic than what's native in Framer. For more complex and detailed prompts I found Workshop just doesn't do well and that's where I switch to Gemini with a custom Gem I put together.


The Gem is trained on Framer's best practices for code components, so it already understands the structure, how property controls need to be set up, and how to write things so that whoever uses the component - myself, a template user or a client - doesn't need to know anything about the code. They just configure it through the UI like any other component.


This custom Gem really opened a door for me that always felt out of reach before, so if you're using Framer I'd definitely recommend it!




Claude + Notion


I have separate Notion pages for each Framer template I'm working on, client projects, and my own website redesign. I use them as simple project management tool where I cross off tasks to myself about where I'm at, what I want to come back to and just keeping track of things.


My problem with Notion has always been the friction of having to actually write and update things. I've been documenting my process because I know it's important, yet it's always been the most tedious thing to do.


The Claude connection to Notion fixed that. Now I just write to Claude - what I finished for the day, what I'm thinking, what I need to follow up on - and it updates the relevant pages for me. Took me less than an hour to set up and I've used Notion every day since.


For client projects it's been helpful for keeping short notes on where things stand. I also store client briefs in there and document my process on current Framer template work, which makes it easier to reflect on what worked, where and when I was slowed by something and what didn't work at all. It all ends up being useful source material for content to post on socials, too.




Flora Fauna


Midjourney is still where I do most of my visual exploration, but Flora Fauna has become the tool I reach for when I want to stay in one place and iterate quickly.


The node-based system is the main reason. Instead of rewriting long prompts from scratch, I'm working visually, branching off from generations I like, adjusting and building on what's already working. Having everything in one place rather than scattered across different chats or folders on my computer makes a real difference when I'm trying to keep a project coherent.


The video generation is worth a mention as well. Mostly useful for when a client project needs some motion or I want to push something further, being able to switch to a better video model from the same workspace - without starting over somewhere else - saves some time and nerves.




The point of all this


The stack will keep changing as these tools evolve, and I'm also aiming to build more custom setups and even small apps for myself based on specific use cases and workflows I haven't fully solved yet.


But the idea behind all of my AI use remains the same - I want to automate all of the mundane, repetitive and boring work, so I can give more time and attention to the creative work that I love doing myself!

G
GeorgiDesigner
Feb 24, 2026