You can now build a fully functional, complex creative tool in less than 24 hours without writing a single line of code.
I just watched an absolutely fascinating breakdown from a tech creator who managed to build a complete manga generation application from scratch in just over 17 hours. The expert admitted right at the start that she has zero drawing skills, her sketches are basically stick figures, yet she wanted to create a professional-looking Japanese comic. Instead of spending years learning to draw, she spent a day building a software tool that draws for her. She calls this process “vibe coding,” and it represents a massive shift in how we approach creative work and software development. By leveraging a tool called Bolt.new, she acted as a project manager rather than a programmer, instructing the AI to build the infrastructure, connect the APIs, and style the front end.
I think this is incredible because it proves that the barrier to entry for building SaaS (Software as a Service) products or custom workflow tools has completely collapsed. The creator didn’t just generate images; she architected a multi-step workflow with distinct features for character consistency, panel layout, and page composition. She walked away with a working web app and a completed manga titled The In Between.
⚡ The Era of “Vibe Coding” and Custom Tooling
The core concept this innovator demonstrates is that we no longer have to wait for companies to build the exact software we need. If you have a specific workflow, like needing to turn bad sketches into consistent manga panels, you can now prompt that software into existence. The creator used a platform that allows users to describe a web app in plain English, and the AI handles the coding, dependency management, and deployment.
She started by asking the AI to plan the build, outlining features like a Character Creator, a Panel Generator, and a Page Composer. The AI even selected the image generation API (which the transcript amusingly refers to as “Nano Banana,” likely a transcription quirk for a specific model or API). The significance here is the speed of iteration. When a feature didn’t work, like the initial API connection failing, she didn’t debug code; she simply pasted the error back into the chat and told the AI to fix it. This “vibe coding” approach allows non-technical creatives to build bespoke engines for their specific artistic needs.
📌 Feature 1: Solving Consistency with Architecture
The biggest challenge in AI art is character consistency, and the creator solved this by building a dedicated “Character Creator” module within her app.
Instead of blindly prompting for a character every time, she built a system where she generates a reference sheet first. The app takes a text description, for example, a Japanese salary woman named Ayaka, and generates front and back views. Once she liked a design, she configured the app to save that image as a “anchor” or reference point.
Then, for the “Panel Generator,” she built a dual-input system. She could upload her terrible hand-drawn sketches (stick figures on napkins) alongside the saved character reference. The AI would then take the composition of the sketch and the style of the character to generate a high-fidelity manga panel. This is a brilliant example of building software to bridge a skill gap. She didn’t just ask for a picture; she built a pipeline that enforces visual rules, allowing her to maintain the identity of her protagonist across different scenes and angles without manual painting.
📌 Feature 2: The “Lazy Mode” and Iterative Prompting
One of the most interesting moments in the video was when the creator realized she could automate even further, moving from a sketch-based workflow to a text-based one.
She asked the AI to build a “Lazy Mode” page generator. This feature accepts a pure text prompt describing a scene and automatically attempts to arrange the panels on a page. While testing this, she discovered that the AI could handle the layout surprisingly well, interpreting instructions like “Tokyo office building at dusk” and placing it in a logical manga sequence.
However, this process wasn’t without friction. She shared a transparent look at the struggles, specifically regarding specific assets. She spent a surprising amount of time trying to generate a consistent image of a “ballpoint pen” character. It highlights a reality of AI development: sometimes the complex code works perfectly on the first try, but the simple creative assets take hours of tweaking. The app allowed her to iterate rapidly, swapping out reference images until the “vibe” was right, proving that these tools are dynamic environments, not just static codebases.
📌 Feature 3: Knowing When to Stop Building
A crucial lesson from this savvy professional was recognizing the limits of her V1 product.
She initially tried to have her app handle the dialogue and speech bubbles. She spent about an hour trying to get the AI to render text directly onto the manga pages. It technically worked, but the user interface was clunky, the bubbles were hard to position, and the font rendering wasn’t crisp.
Instead of sinking another 12 hours into coding a complex drag-and-drop text editor, she made a smart product decision: she exported the clean images from her app and moved them into Canva to add the text. This is a vital takeaway for anyone looking to build their own tools. You don’t need to reinvent the wheel. The goal of “vibe coding” is to solve the specific problem that doesn’t have a solution (generating consistent manga from sketches) and then rely on existing robust tools (like Canva) for the standard stuff. This hybrid workflow allowed her to finish the project in under 24 hours rather than getting stuck in development hell.
This video is a must-watch if you want to see the future of creative workflows!