Inkwell: Building Apps with Sub-second Image Generation

Modular co-founder and president Tim Davis built Inkwell, an AI-powered children's storybook app, using Modular's Flux and Gemma 4 endpoints. In this demo, he walks through how the app generates stories and images in real time, including choose-your-own-adventure branching, character consistency across scenes, TTS narration, and content moderation via a second LLM. The conversation covers what made sub-second image generation achievable (Modular's vertically integrated stack from kernel to cloud), the UX decisions Tim made to handle latency, and where else fast image generation is useful beyond storybooks. If you want to build something similar, Modular's Flux endpoint is live at modular.com with a playground on the homepage. 0:00 Intro 1:52 What is Inkwell 3:08 Live walkthrough 5:49 Image generation speed 8:39 Writing your own story 11:08 Story creation demo 12:15 Other use cases 13:25 What Modular Cloud provided 16:39 How Modular's stack makes this possible 18:51 How to get started 20:35 Closing