Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
April 25, 2025
·
Singapore
Stable Diffusion LoRA Portraits
Learn how to fine‑tune Stable Diffusion with LoRA using personal photos, then generate on‑brand portraits for social media without costly shoots.
Overview
This project explores how solopreneurs can leverage generative AI—specifically LoRA (Low-Rank Adaptation)—to create personal, on-brand photos for social media, without the need for expensive photoshoots. By fine-tuning a model on my own images, I aimed to generate scroll-stopping, contextually relevant portraits that evolve with my brand.
Technical Approach
- I used a dataset of curated personal images (front-facing, well-lit, varied poses) to fine-tune a Stable Diffusion model using LoRA.
- Training was conducted via Replicate LoRA Trainer.
- I’ll walk through the data gathering process, training process, and results.
- I’ll also share my prompt generation using ChatGPT 4o.
Links
Trained a GenAI LoRa model on selfies to generate realistic images.
Tech stack