Building And Running An AI-Powered Product
summary
A hands-on runthrough of building Storique: a private‑beta personalized children’s book app that actually works in Switzerland. It fine‑tunes Stable Diffusion variants using stable diffusion XL and LoRA fine‑tuning, trains a user avatar from about eight photos (training ≈30 minutes), reduces inference to FP16 and ~20 steps (was 50) for cost, and uses Hugging Face diffusers with Replicate for deployment plus LangChain and GPT‑3.5 for text; result: 24–36 page books generated in roughly half an hour.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:25AI product development today
- 3:10Storic and AI storybooks
- 4:13Stable Diffusion for illustrations
- 9:55Fine-tuning Stable Diffusion
- 16:30Optimizing Stable Diffusion
- 23:04Deploying models
- 27:49Live demo
- 37:00Key takeaways
- 38:26Q&A
related talks
36mDesigning the Future: Stable Diffusion & the Rise of AI-Driven SaaS
34mBuilding Owly: An AI Comic Video Generator for My Son
19mAI-Powered Startup Toolkit for a Product Designer
22mBuilding next-gen speech synthesis for Bookmate audiobook service
21mBeyond the Chatbox: What Changes When AI Runs Where You Are
42mCreating Smarter Applications
45mBridging the gap between AI Research and Human-Centered Products
28mComputing Large Foundational Models