CineLeopard macOS Video Editor
CineLeopard is a native macOS video editor I built with SwiftUI. It is available now on the Mac App Store. I wanted a focused, minimalist editing experience and the CineLeopard app provides that.
I design custom apps and software.
- TypeScript / React web apps and websites
- Custom AI integrations
- Custom iOS apps and macOS software
- Experienced builder and designer
Available for new projects
cortney@cortneyharrington.comWork
A selection of the apps, software, websites, and AI systems I've designed and built — examples of what I can build for your business.
Video Production Planning App
A TypeScript web app for planning video productions end to end. It helps organize the moving pieces of a project. The screenshot below shows the planner being used to map out the making of my Superhighway video. This is currently an internal tool. I will ship a beta version of this project soon.
CineLeopard Video Production Company
The website for my video production company, built as a TypeScript site. It serves as the home for CineLeopard's video production work.
SimuLeopard Version 2
SimuLeopard now encompasses a wide array of engineering design and intelligence tools. In addition to its original simulation functionality, SimuLeopard helps users create first principles engineering designs, conduct agentic first principles materials research, engage in agentic manufacturing planning and sourcing workflows, and produce a wide range of immersive creative media. SimuLeopard's rendering capabilities let you create product and machine images in many popular industrial design styles. In addition to simulations and physics-based analysis, SimuLeopard gives builders and makers concrete starting points for prototyping, accelerating the process of moving from ideas to atoms.
SimuLeopard
I am working on a new project called SimuLeopard. It is a web app that lets you create code-based engineering, manufacturing, and physics simulations from text, image, and CAD prompts and visualize your simulations with technical images and cinematic videos.
Module Dog
This is my first AI education project. Module Dog is a web app that lets teachers create engaging and rigorous lesson plans and learning materials using AI. Teachers can save their specifications for each class, quickly generate lesson materials like texts and handouts, search for images and texts to add to their modules, and even export their lessons as formatted PDFs and slide presentations. Many of the teachers I work with have expressed their dissatisfaction with some of the online curriculum platforms they are encouraged to use. When I was a multi-subject middle school teacher, the lesson planning involved was extremely time-consuming. Module Dog would have helped me tremendously. I also wanted to make something that would address the real issues I see in high school classrooms involving lack of engagement, lack of rigor, and lack of substantial reading and writing practice. Module Dog is meant to be simple, straightforward, and effective. It lets you do what is best for the students quickly and easily. I built Module Dog with Next.js/React, Node, Supabase, Stripe, and various other APIs. I wanted to make something that could help me become more skilled at building complex databases and Module Dog has been that project. I also created a workforce training version of this app and can create similar custom projects that automate course creation for schools and businesses by ingesting text and multimedia content.
Craft Rabbit
This iOS app lets users create craft, artisanal baking, interior decor, and architectural images. Users can then get text instructions to make what is in their image. They can also upload their own images to ask questions and get instructions. I like AI projects that connect users back to making physical things in the real world. I built this with Swift UI in Xcode. Some of the image models are my custom fine-tunes. It also uses a couple of different multi-modal models for the text generation. I built Craft Rabbit in summer of 2024. I added the Prompt Palette in November 2024. The Prompt Palette offers users a unique way to make image prompts by tapping on a keyboard of craft and decor items, including many possible choices for styles, colors, materials and patterns. You don't have to think hard and compose detailed descriptions to be creative.
Modern Art AI
Modern Art AI is my fashion design web app. I used Next.js/React and APIs connecting to my custom fashion image generation and sewing-pattern models. The UI also lets users upload their own sketches and images to guide their creations.
AI & Model Training
Pattern Design Models
Modern Art AI started out as a sustainable automated textile-manufacturing concept. I started learning about and experimenting with AI in order to develop an image-to-sewing pattern system that could drive the customization process for garment creation. I made some datasets of hand-drawn vector image sewing patterns derived from real clothing images (image-to-image training pairs) that were calibrated based on real-world physics. While iterating on these datasets, developing pipelines, and training several kinds of diffusion models, I eventually reached a temporary solution that involved blending an input image into the image generation inference processs. While this produces some interesting patterns that can be scaled to working size, printed, and used, I want to build a more grounded image-to-image pipeline that also has a model trained on the exact geometry of the patterns and clothing. I've learned a lot since I started working on this. I plan to revisit this problem again soon.
Image Generation Models
I started training some image generation models in Fall 2023. Many of the image datasets I created as a basis for my pattern-generation vector drawings were begging to be trained as straightforward text-to-image diffusion models. I built increasingly larger and better-curated datasets over the winter of 2023-2024. I did some lighter (LORA) fine-tunings of SDXL and some deeper full fine-tunings. I experimented with many different datasets and types of images (runway versus editorial, commercial contemporary design versus more timeless or experimental assortments). I wanted to override some of the aesthetics of the base models with my taste, with an emphasis on encoding fashion and aesthetic terminology very precisely for better instruction-following. I also trained some more generally decorative models with a mix of architectural, interior design, and crafting images. Two of those models became part of my Craft Rabbit iOS app.