Independent Product·Web App·2024
SlideWithAI
AI-generated decks with voice narration, captions, and embedded quizzes.
Overview
SlideWithAI turns a prompt into a polished, narrated, interactive deck — layouts, content, voice, and engagement included. The product is built for people who lose hours every week to slides that still look mediocre, and for talks that need to do work when the presenter isn't in the room.
The problem
Slide creation is one of the most universally tedious parts of knowledge work. Existing tools either ask too much (manual layout, content, design) or generate output that's structurally fine but tonally lifeless. And once the deck is built, it remains static — no narration, no checks for understanding, no signal that anyone actually engaged with it.
"Stop spending hours on slides that still look mediocre."
Approach
The product breaks down into four interlocking surfaces. Each one had to feel coherent on its own and seamless together — generate, edit, narrate, engage.
- Generation engine. A prompt-first flow that picks layout, density, and tone from 52+ professional templates. The hardest design problem was disambiguation moments — making the model ask the right question instead of guessing.
- Voice and captions. Natural narration with word-by-word caption sync. Designing "regenerate just this paragraph" so it felt safe and predictable was the trickiest UX problem.
- Embedded engagement. Quizzes and polls authored inline alongside slide content, with preview affordances that show the viewer experience before publishing.
- MCP and export. One-click PDF export, share links with view analytics, and an MCP integration so any AI assistant can spin up a deck on demand.
Outcome
SlideWithAI is live with a generation-to-share flow that takes minutes, not hours. The template system spans 52+ layouts and 5 aspect ratios across teaching, sales, and creator use cases. Decks "talk back" through narration plus caption sync — useful for self-paced learning and async sales — and the MCP integration means any AI assistant can produce a finished deck in a single conversation.
Reflections
- Systems thinking matters. Building a working AI product meant designing the whole loop — generation, editing, narration, sharing, and analytics — not just a clever prompt interface.
- Design tools are still needed. Coding fast is powerful, but maintaining a coherent design system purely in code gets difficult. Tokens, reusable patterns, and visual QA still matter.
- Builders should learn to sell. Building has become easier, but getting a product to launch, positioning it clearly, and earning attention is still real work.
Want a closer look at the generation flow, narration UX, or MCP integration?
I'm happy to walk through the design decisions behind each surface.
Try it at slidewithai.com →