Refs.design

• 

  Generate professional grade AI imagery

• 

  Save your references. Recreate any image.

Save your AI image references in one place.

Generate professional AI image prompts in seconds.

I've struggled to create great AI imagery for the longest time. So I built something to fix it.

After struggling to create great imagery myself using image generation tools, and also using AI image generators more and more for creative projects for my design agency, I wanted to create a simple platform that solved 2 clear problems: a single place to store all of my image references, a place to generate professional grade image prompts to use in image generators.

I also saw an opportunity here as this is something a lot of designers struggle with, and then went through my own startup validation method and prompt process to validate the idea before building.

What it does

Essentially, refs.design allows you to upload all of your image references and inspiration so you can use them super quickly in image generation tools like midjourney. It allows you to generate new image prompts from any image, as well as generate custom image prompts from scratch using my own custom 'AI Creative Director' system prompt.

The key features are:

• Upload any image using copy/paste
• Generate style prompts from any image in one click
• Copy any image to clipboard in one click
• Generate professional prompts using the prompt builder
• Store all your AI image references in one place

How I built it

I built it over the course of 3 sessions using Bolt, Supabase, Netlify, Github, Make.com, Midjourney and OpenAI API. I also used Figma to design initial screens for the website pages and the dashboard.

First, I validated my idea using my custom startup validation prompt (watch my Youtube channel for this one!), mapped out the user journey using Whimsical, then scaffolded the UI from Figma in Bolt. Then I added the onboarding flow and basic functionality. Then I added in the prompt generator. Then I added the finishing touches and fixed any bugs before finally deploying using Github and Netlify.

This was all built in the space of about 3 days over the course of the hackathon, as a passion project really to help me with something I struggle with which is getting great images out of image generation tools.

Challenges I ran into

There were some technical challenges such as setting up the automations correctly in Make, and also having to cut back scope before the deadline for paid plans which I was planning to integrate with Polar.

I also ran into some DNS challenges with Netlify specifically around deploying single page apps and deploying to subdomains but thankfully all resolved before submission.

Accomplishments that I'm proud of

I'm mainly just proud of building a tool I can use on a day to day basis, and also for being able to showcase the process from start to finish on my Youtube channel for my audience as well, who I'm helping to teach how to build these apps from scratch.

It's not the most ambitious project, but it is a simple app, that I'm gunna use every day, solves a problem I have (and I know other designers do too), and also just gives me one place to store all my inspiration that's far better than a folder on my desktop!

What I learned

I've learned so much in the process of building apps and deploying them that I never knew before, or only knew the intricacies of before starting.

I've helped startups design products for the last 15 years and have been designing them through my own agency for the last 4 and I'm now able to not only design but also build and design in code for clients which is incredible.

What's next for Refs.design

Next the plan is to validate demand, test social ads and organic audience building for the product on short form video using AI tools to generate ads in repeatable formats (thanks Greg Isenberg for the video on that one).

If demand is there and people sign up to the product and waitlist for the paid version, we'll see about adding the paid plan and then just see how it goes and develop it with audience and users in mind first. I've got some ideas but I know that listening to users and building with community and audience is key.

I'm also going to be pulling this into Cursor + Claude Code (all the current code is 100% written by Bolt) to flesh out a broader project plan for additional features and ensure all the code is 100% secure.

Looking forward to seeing where this goes. (And I've got plenty of other products in mind if this doesn't work out!)

Built With

• bolt.new
• make.com
• netlify
• next.js
• openai
• react
• supabase