Hello Product Hunt community!
I m Kensu, a professional interior designer. Over the past year, I ve been fascinated by generative AI's potential in spatial design. However, like many in our industry, I kept running into one massive roadblock: AI simply hates architectural precision.
In a professional workflow, you can't have the AI warping the building geometry, changing the camera height, or shifting the perspective. Every pixel counts when delivering to a client.
That s why, in my spare time, I decided to build KENSU-VisionFlow (https://visionflow.kensustudio.com).
I wanted to create something that enforces a strict "Camera Lock" (ensuring the output photo is pixel-identical in camera perspective to the base drawing) and extracts "Lighting DNA" from reference images without messing up the layout.
We are officially launching tomorrow (July 13)! As we prepare for the big day, I d love to open up this forum and hear from you:
Have you tried using AI in your design or 3D rendering workflow? What is your biggest frustration?
If you're curious about our tool, leave a comment with your thoughts or questions below, and I ll gladly send you some free trial credits to test it out before we go live tomorrow!
Thank you so much for your support, and I look forward to chatting with you!
Hi 👋 I’m xiaohui, a professional interior designer turned indie hacker, and I’m incredibly excited to share VisionFlow Pro with you today.
The Problem:
Like many of you, I was amazed by AI rendering tools at first. But when I tried using them for actual client work, the frustration began. Using AI felt like playing a random "gacha game." You upload a clean SketchUp model, and the AI hallucinates fake LED strips on flat walls and completely ruins the ceiling structure. It lacked the precision required for professional architecture and interior design.
The Solution:
I realized we didn't need another generic image generator; we needed a strict, predictable rendering engine. So, I spent my spare time building VisionFlow Pro.
Here is what makes it different:
Dual-Agent Lighting Architecture: I built a custom routing system that extracts the abstract "mood DNA" from a reference image and maps it to your base geometry. It strictly locks your structure—meaning zero hallucinated windows or fake light fixtures.
Traceable Asset Hub: I built a robust file management system. Every base image, reference, mask, and result is archived into dedicated project folders with a full version history. It makes AI traceable, reversible, and reusable.
Bring Your Own Key (BYO-Key): I don't mark up GPU costs. You plug in your own Google Gemini API key and pay raw compute costs (fractions of a cent per render) directly to Google. You buy the software lifetime license once, and that's it.
The Journey:
While building this, the hardest part was taming the AI's bias to "invent" lighting. It took countless iterations to develop strict rendering protocols to force the AI to respect the laws of physics and architectural design.
Onboarding & Free Trial:
Since VisionFlow Pro is built for professional workflows and utilizes your own Google API Key, I am personally onboarding our first batch of users via email. Click the link to apply for a 7-day free trial. I’ll personally send you the package and help you configure everything for free to make sure you get perfect renders.
I would love to hear your feedback, answer any questions about the technical architecture, or hear about your rendering workflows. I’ll be hanging out in the comments all day!
Thank you!