In-Depth Review: LookX AI
What It Is
A browser AI image tool for architecture that turns a text prompt, a sketch, a photo, or a screenshot of a model viewport into rendered images, video, and upscales, with plugins for SketchUp and Rhino and the option to train a model on your own images.
Who It's For
Architects and designers doing concept and mood work who want a house style the tool can learn, rather than prompting from scratch each time.
Why It Matters
You upload a set of images that share a look and the tool fine-tunes a model on them, so a studio can generate in its own visual language rather than prompting for it each time. Few tools in this category offer per-firm training.
Problem Solved
Generic image generators do not know what a plan, a section, or a curtain wall is, so the prompt does the work. LookX is trained on an architectural dataset the company calls ArchiNet, which it says carries building function, form, view, and style as structured annotations. A Style Adapter matches a reference image, and a vocabulary library supplies architectural terms so you are not inventing prompt language.
Audience Fit
Primary Users:
- Architect
- Designer
- Student
- Visualization Artist
Ideal Fit:
- Concept and mood studies where a look matters more than accuracy
- Studios wanting a model trained on their own past work
- SketchUp or Rhino users who want generation near the model
- Anyone testing AI imagery at $20 per month before committing further
Key Differentiator
Trains a model on your own images so a studio look becomes repeatable
Workflow Fit
Project Phases Covered:
- Concept
- Schematic
Where It Plugs In:
Browser tool with SketchUp and Rhino plugins. Images in and images out, with no geometry, BIM data, or documentation produced.
User Experience
Speed: Cloud generation. A paid plan makes image generation unmetered and puts you in a priority queue
Interface: Render and Explore modes, a Style Adapter for reference matching, an upscaler, and a vocabulary template library
Learning Curve: Gentle to start. Practitioners report the difficulty is control rather than operation, since repeating a result is unreliable
Collaboration: Team plan covers five seats with a shared point pool, a team cloud drive, and team-only models
Integrations: SketchUp and Rhino plugins. An API is offered on the enterprise plan
Bonus Insight
Two practitioner reviews reached the same conclusion and they are the most useful thing written about this tool. Both found it produces striking images and neither could make it repeatable: seed locking does not reliably reproduce an output, the hue-imitation setting drifts from the source geometry, a reference image can transfer its colors literally onto materials, and the tool sometimes changes the design while improving the picture. Architectural work needs revision, so treat this as a concept tool rather than a client-deliverable one. Some practical cautions. The site's HTTPS endpoint was failing to complete a handshake when checked on 2026-09-13, so a browser that forces HTTPS may show nothing. The pages load over plain HTTP. The company was founded in Los Angeles in 2023 by Wanyu He, a former OMA architect who also co-founded XKool Technology in Shenzhen, and the hosting still runs on Chinese infrastructure. No funding, team size, or customer has been disclosed for the US entity, and the real press coverage is all from 2023. There is no published privacy policy detail, no retention terms, and no statement about whether uploads are used for training, which matters if you are uploading client imagery.
Need help implementing LookX AI?
Our advisory team has hands-on experience across 550+ AEC tools. Book a free AI strategy session to get specific recommendations for your firm.