Realtime AI Rendering
Like Nano Banana 2 but much faster.
Shaun McCallum
March 30, 2026Recently we teased realtime ai rendering and a lot of you signed up for access so you’re first to hear about this. If you have used Nano Banana 2, you already understand how powerful AI rendering and editing images can be. It’s like that, but much faster.
What we have been building takes that same idea, but brings it into realtime.
You can now edit images, 3D models, and styles instantly with simple prompts. Instead of generating, waiting, and retrying, you stay in flow and iterate continuously as you design.
It works best when you start with a solid sketch or 3D model. From there you can explore materials, lighting, environments, and even weather conditions in milliseconds, making it much easier to test ideas and move quickly.
We have also introduced a set of built-in styles that are one click away while in realtime:
- Photorealistic outputs with no prompts
- Vector styles
- Focused concept visuals
- Sketch styles
- Anime and manga
Switching between these is instant, so you can compare directions and make decisions without slowing down.
From today, realtime is available to all subscribers in alpha (https://alpha.fenestra.app).
Over the next week, we will be transitioning everyone over to the alpha user interface, so expect rapid improvements as we continue refining it
Give it a try and let us know what you think. What works, what doesn’t and what you want to see next.
Try Realtime Today!
Render Now!Related Articles
Fenestra 101: Our First Webinar
We're running our first ever live webinar on June 3rd. A hands-on intro to Fenestra for architects and interior designers, covering image generation from prompts, 2D and 3D inputs, video, and where AI actually fits in your practice. 30 seats only.
AI will design all architecture.
A step-by-step look at using GPT-Image-2 and Fenestra to design a Tokyo-inspired residence — from generating the base image to material studies and a full competition board. Simple prompts, real results.
GPT-Image-2 is a serious contender for best editing model
Testing GPT-Image-2 on a Japanese-inspired teahouse: prompt-based edits, expanding a single image into a full competition board with plans, sections and elevations, and generating creative multi-view triptychs in a single generation.
