Overview and First Impressions
Upon visiting Amazing Photos at https://amazing.photos/, I immediately noticed the developer’s personal touch. The landing page is a first-person blog post by Tom Dekan, not a polished marketing site. That honesty sets the tone: this is an open-source tool built by a solo engineer, not a VC-funded startup. The home page doesn't even have a typical dashboard; instead, it leads you to a GitHub repository and a sign-in prompt via Google OAuth. Once I logged in, the clean interface of a Next.js app appeared: a simple upload form, a prompt box, and a gallery of generated images. The app is intentionally minimal—no dark mode toggle, no advanced settings panel. The focus is entirely on uploading photos and generating new ones with custom prompts.
How It Works: Training and Generation
Dekan’s workflow is straightforward and well-documented. After signing in with Google, I uploaded 20 high-resolution shots of my face (the recommended amount). Behind the scenes, the app sends these images to Replicate’s Flux fast trainer—an H100 GPU-backed model that finishes tuning in about two minutes. I received an email notification with 12 sample images generated from my trained model, all rendered in various styles (photorealistic, cartoonish, even “as a taco” as a fun test). From there, I could type any prompt into the text box and wait roughly 20 seconds for a single output. The app currently generates images synchronously, meaning you cannot queue multiple prompts at once. This feels glacial compared to competitors like Leonardo AI or Midjourney, which produce multiple variations per request. However, the uncensored nature is a genuine differentiator: no filter blocks creative or adult prompts, as long as you adhere to basic decency. The open-source (MIT license) aspect is also rare—you can fork the entire repository, run it locally, and modify the stack as you see fit. This makes Amazing Photos a fantastic learning resource for developers wanting to understand how to build an AI photo generator from scratch.
Performance, Tech Stack, and Developer Experience
Dekan is refreshingly transparent about the technology. The stack includes Next.js (App Router), TypeScript, Tailwind CSS, Better Auth, Stripe, Prisma, and Vercel for hosting. He openly criticizes Next.js for slow hot-module reloading in larger codebases—a pain I’ve experienced myself. During my tests, the app’s response felt snappy on the front end, but the 20‑second generation time (due to sync calls to Replicate) quickly frustrated me. I also noticed that the training step consumes GPU credits on Replicate, and the app uses Stripe for payment, but pricing is not publicly listed on the website. The only clear free offering is the demo that uses Dekan’s own trained model (you get to see yourself as a taco, but not your own face). For users who want to train their own models, the cost is hidden until you attempt to generate more than the free samples. This lack of upfront pricing is a significant hurdle for casual users.
Verdict: Who Should Try Amazing Photos?
Despite its rough edges, Amazing Photos fills a niche. Developers or AI enthusiasts who want to understand the end-to-end pipeline of training Flux models via Replicate will find the open-source codebase invaluable. The uncensored policy and fast training time are real advantages over services like Midjourney or Leonardo AI, which often impose content restrictions and need longer training queues. However, non-technical users expecting a polished, hassle-free tool should look elsewhere—the synchronous generation, minimal UI, and unclear pricing are deal-breakers. Strengths include full transparency, MIT licensing, and the educational value of reading a developer’s honest postmortem. Limitations include the lack of parallel generation, poor documentation for end users, and the reliance on a single developer’s maintenance. I recommend Amazing Photos to tinkerers and indie developers who enjoy self-hosting and experimenting. For a quick, reliable, and feature-rich image generator, stick with the big-name platforms for now.
Visit Amazing Photos at https://amazing.photos/ to explore it yourself.
Comments