Hey guys 👋, Syakir here!

Its been a while since my last post. I’m currently busy building app as an indiehacker. Now, i want to share about my latest project, Kattalog, an AI product photography SaaS.

This project is not the first of its kind, since there’s already some similar tools out there. But i’m still building it anyway for these reasons:

• An experiment & learning journey to become an AI engineer & SaaS builder
• The app target market are ecommerce sellers, and I worked in Ecommerce industry for the last 5 years
• Based on the initial research, the market is not saturated yet
• The competitor apps are mostly very technical (require detailed prompts), and i try to fix this with different approach (image to image)

How I Built Kattalog

The story begins about 3 months ago in the end of February 2025, when i had a chat with ChatGPT. I was chating about building another directory-site, like AlternateOSS, but for AI generated images. Initially, i was going to build an AI stock photos website.

I researched about it, also with ChatGPT, from the AI models to market competitions, and realize that it’s not the type of product that i can build solo. Then, i ask ChatGPT to give me other ideas that fit my contraints, but still related to AI images with not-really saturated market. That’s when i decide to built Kattalog, an AI product photography app.

Then, i tried the existing competitors to find my app value proposition. They gave me a lot of idea, but in the end i only focus to build only core features for MVP.

I started building it since early March 2025 and the MVP done and deployed by the beginning of this month, May 2025. Initially, i vibe-coded this project using Cursor. About 50% codes are generated by Cursor. I also generate and customize most of React components using v0.

I’m building it while having a full-time job, mostly in the morning before work (I’m an early bird). Cursor and v0 really helps alot here. They make the development significantly faster. Although, i still need to code manually for specific cases.

The main challenges when developing Kattalog is actually not coding, but finding the right AI models to use. I literally spend about 3 weeks researching about them.

I can’t disclose which AI models i use for Kattalog, but i use several models from Replicate & Fal.ai. I didnt fine-tune or train my AI models for now, but i plan to do it later if it get some tractions.

That’s why i said “It’s not as hard as you think” in the title. lol.

You can also follow this approach to build your own SaaS. In fact, most AI Saas today use the publicly available models. So, start building your own AI project. It will be even easier if you use Cursor, even the free trial. It will help you a lot developing the prototype.

How Kattalog Works

The main functionality of Kattalog is replacing the product image background with studio-quality background with AI. Instead of doing expensive product photoshots, or photoediting, ecommerce sellers can just upload their product photo, and let Kattalog do the works.

Here are the high level flows:

• User upload product images
• Kattalog will process and remove the background
• User open the product image with removed background in the editor
• User then choose a background reference, and click generate
• 4 product images will be generated with new backgrounds.

You can find real examples of generated images here: https://kattalog.com/examples.

The main feature (background generation) is only available in paid plan. As a solo founder, i cant afford to make it free, since it cost me money to run it.

But if you’re an ecommerce seller and wanted to try it, i can give you free access. Just let me know.

Apart from the paid feature, there are also free features available to use:

• Background colors
• Pre-generated backgrounds

The Stacks

The stacks i used are simple. I use my favorite framework Astro, but now in SSR mode

The Frontend:

• ✅ Astro
• ✅ React
• ✅ TailwindCSS
• ✅ Shadcn / Radix
• ✅ Konva.js for Image Editor

The Backend

• ✅ Astro SSR APIs
• ✅ PostgreSQL
• ✅ Drizzle
• ✅ BetterAuth
• ✅ Cloudlfare R2 for storage
• ✅ VPS & Self-hosted Coolify for deployment
• ✅ some AI models from Replicate & Fal.ai

I use Polar.sh for payment processor, since Stripe doesnt support my country (Indonesia)

The Real Challenges

The real challenges of building SaaS is not product development. It’s product market fit. And, this is where i’m, looking for my first customers.

To be fair, i didn’t market it yet extensively. But, I started it today by publishing it here. Tomorrow, i will also launch Kattalog in Product Hunt.

Wish me luck! :D

—

What do you think about this project?

If you have any feedback, please dont hesitate to put it in the comment below.

Thanks, happy building!