What’s your biggest hurdle with AI rendering?
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!
Replies
To kick off the discussion, here are the 3 biggest hurdles I’ve personally struggled with when using generic AI for interior design, and what we set out to solve:
1. Strict Geometric & Structural Fidelity (The "Camera Lock" Problem):
Most AI generators love to "redesign" the space. But for professional delivery, we need the AI to treat our base drawing as an absolute boundary. The camera position, ceiling height, and wall layouts must remain pixel-identical, not re-imagined.
2. Decoupling Lighting from Style Pollution:
Often, we find a reference photo with gorgeous lighting, but we don't want its furniture or materials. Decoupling is incredibly hard—how do we extract only the "Lighting DNA" (color temperature, shadow softness, direction) and inject it into our design, while discarding the reference photo's actual geometry and objects?
3. Multi-Angle Spatial Consistency:
When presenting a concept to a client, we rarely show just one image. If we generate 3 different angles of the same living room, how do we ensure the sofa texture, the flooring wood grain, and the spatial context remain perfectly consistent across all renders?
I’d love to hear your thoughts. Which of these three is the biggest headache in your daily workflow? Or is there a 4th hurdle I missed? Let's chat!