I built a one-image workflow for brick model build steps.
The interesting constraint is that the input is only one image.
That makes the product different from CAD tools or manual instruction builders. It is not trying to recover a perfect hidden part list from a photo. Instead, it creates a visual explanation layer that is useful for demos, launch pages, product listings, and creator posts.
Workflow
- Start with one clear finished model photo or render.
- Generate a staged build-steps board.
- Review whether the sequence is readable.
- Render a short front-view assembly video from the same source workflow.
The current public examples are here:
https://brickgpt.online/examples
What Makes A Good Input
The best input image is not always the prettiest image. It is the image with the clearest structure.
Good inputs usually have:
- a centered subject,
- stable lighting,
- visible model edges,
- a quiet background,
- and a front or slight three-quarter camera angle.
Bad inputs include cluttered desk photos, strong reflections, hands blocking the model, and images where the model is too small in the frame.
Why This Exists
A lot of brick model content shows only the finished object. That is fine for visual inspiration, but it does not explain the build.
The steps board gives the viewer an intermediate explanation. The short video makes that explanation easier to share.
This is especially useful when the goal is not engineering precision, but communication.
Clear Limits
BrickGPT is independent and not affiliated with LEGO.
It should not be used as official replacement documentation. It also does not claim to create CAD-accurate instructions from a single photo.
The intended use case is presentation: one finished model image becomes a clearer build story.
Project link:
Feedback I am looking for:
- Does the board make the build progression understandable?
- Would the short video help on a product page?
- What kind of input image would you test first?
Top comments (0)