A developer builds a virtual try-on feature from scratch using Kilo Code, an AI coding assistant. They scope requirements, generate a design doc with the AI, then code and debug through five iterations until the animated avatar works with a video-generation API. The whole process takes one walkthrough.
How would it be to code a new feature for my app where I show you my full process of building software with some of the most powerful AI tools to also help you turn your idea into reality? Today we're integrating a new API to our virtual try-on-crop extension. Currently, the output is a static image, but I got a few suggestions to make the avatar move, which requires us to integrate a video generation API. But before I write a single line of code, the first step is always to scope out the product requirements and user flow, which I then translate into a technical design doc using Kilo Code's architect mode. This is a new open-source AI coding agent I'm trying out, built by the founder of GitLab. I'm honestly shocked at how good it is. Rather than the slop that most LLMs write, Kilo not only was able to proactively ask me questions to resolve ambiguities in the design. It also was able to spin up a browser agent to visit Gemini's API page before generating the final design doc. I then turned on coding mode to implement the various components outlined in the doc, ran the extension, got some errors, and placed them on debug mode. After like 5 rounds of debugging and prompt training, the avatar moved! Now you can see the trial result from front to back. At this speed, I plan to launch the extension in the next 21 days. So comment Kilo if you want to try my app, or use Kilo Code to build your own.
