The video is the beginning. The model is the foundation.
A video shows one explanation of an assembly. A structured model can become the foundation for many explanations, each suited to a different person and task.
One product, several questions.
A buyer wants to see how the product comes together. A technician wants to know the next step. A documentation team wants to update every instruction affected by a component change. An engineer may want to investigate whether the proposed order leaves a fastener accessible.
Those questions share underlying information: the parts, their relationships, the operations and the dependencies between them. Our direction is to make that information reusable.
What an assembly graph would add.
The current engine stores parts and supported placements. A richer assembly graph would also express which components connect, which fastener joins them, what must happen first, and where an interpretation needs review.
Consider a cover secured with four screws. A graph could connect the cover to the housing, associate each screw with its hole, and record that the internal module must be fitted before the cover. That relationship could inform both an animation and a step-by-step guide.
Several outputs from the same understanding.
- A video for a product page or sales conversation.
- An interactive guide that lets someone inspect one step at a time.
- A documentation export linked to the component inventory.
- An integration that carries approved assembly information into another workflow.
Video is the working output today. Interactive guides, richer graph relationships and customer integrations are planned work.
Human corrections should remain visible.
An expert may correct a dimension, identify a missing fastener or change the order. Our intended review workflow should preserve those decisions and their source, rather than hide them behind a finished animation.
Uncertainty also needs a useful explanation. “This orientation needs review because the drawing has no side view” is more actionable than an unexplained confidence percentage. Numerical confidence would need evaluation and calibration before it could support a decision.
Build the foundation in steps.
We are starting with clear assembly videos, then moving toward better review tools, an assembly graph and interactive delivery. The longer-term question is whether that understanding can help people identify assembly problems earlier.