Every connection. In order.
A 46-part pump housing, built one placement at a time. Follow the rotor, cover, bearing cap, coupling and side ports through to the finished assembly.
Every product has a way it comes together. Latch makes it visible, starting with assembly videos, building toward a deeper understanding of the physical world.
Turn supported shop drawings and illustrated PDF guides into assembly concept videos. Help customers, partners and teams see what a product is made of, and how its components come together.
The AI interprets the documentation. A structured model defines the geometry and movement. You review the result against the source.
See what the engine can makeStart with a readable PDF, known hardware and assembly context.
Translate the source into components, geometry and supported placements.
Check source notes, assumptions and the sequence. Revise what needs work.
Create a labeled MP4 in landscape, square or portrait format.
A pump housing and a workshop cabinet. Follow the connections one at a time, then see the complete assembly.
A 46-part pump housing, built one placement at a time. Follow the rotor, cover, bearing cap, coupling and side ports through to the finished assembly.
Shelf supports, six shelves and a rear panel complete a 61-part workshop cabinet. Close-up cameras follow each connection before returning to the finished build.
Rendered from original synthetic drawing studies, with reviewed assembly plans. These are illustrative guides, not customer projects or unattended PDF-extraction benchmarks. Threads, torque, tolerances and physical performance are not simulated.
Latch is building toward an assembly intelligence platform: a shared model of parts, relationships and sequence that can power videos, interactive instructions and connected workflows.
Assembly Animator is the starting point. The next chapters build on the information behind the animation.
Start with a video. Make the model easier to review. Then use that shared understanding to guide the build, connect teams and investigate problems earlier.
Bring the build to life.
Turn supported PDF drawings into reviewable 3D assembly videos. This is the first working product and the starting point for everything that follows.
See it. Check it. Refine it.
A visual review workspace for comparing the model with the drawing, resolving uncertain details and refining the sequence before sharing it.
Connect every part of the picture.
Build a reusable assembly graph that connects parts, fasteners and dependencies. Expand the foundation toward native CAD and bill-of-materials inputs.
A clear path through every build.
Turn the assembly model into interactive instructions people can follow at their own pace, from a product page to the workshop floor.
Get assembly knowledge where it belongs.
Connect reviewed assembly information with the systems teams already use for engineering, documentation and manufacturing.
Spot the problems before the build.
Explore whether the assembly model can surface sequence conflicts, hard-to-reach fasteners and buildability issues earlier in the process.
Assembly Animator works today. The other tool names are working names for the roadmap; their features are not available yet. The sequence shows our direction, with no promised release dates. Sentinel remains exploratory, not engineering certification.
The process behind the product, the questions worth asking, and where assembly intelligence could go next.
What happens between a PDF and the finished animation, and what makes a drawing useful input.
Read the articleWhy Latch is building toward a shared understanding of parts, relationships and sequence.
Read the articleA practical review checklist for AI-assisted assembly videos: source fidelity, hardware, sequence and limits.
Read the articleLatch is building software that understands how physical products go together. Assembly Animator is our first product: it turns supported PDF drawings and assembly context into reviewable 3D assembly concept videos.
It interprets supported PDF drawings, creates a structured component model, plans supported placements, and renders an MP4 with labeled assembly steps. The workflow keeps assumptions and source notes available for review and supports revisions.
The current drawing workflow accepts PDFs, including shop drawings and illustrated instruction guides. Each file can contain up to 20 pages and be up to 8 MB. Native CAD formats such as STEP, IGES and DWG, plus BOM-linked ingestion, are planned rather than available today.
The AI supplies a structured interpretation. Geometry and motion are then rendered from that model. Parts persist across frames, and the saved plan can be checked and revised. The interpretation still needs to be compared with the source drawing.
Yes. The pump and cabinet videos were rendered by the current engine from original synthetic drawing studies and reviewed assembly plans. They show sequential placements, connection-facing cameras and the completed assembly. They are not customer projects or proof of unattended extraction from unseen PDFs.
No. Current checks can flag modeled solid overlaps and blocked straight placement paths. They do not establish source accuracy, physical fit, tool access, torque or structural performance. Broader assembly validation is a future research direction.
Yes. The local CLI prepares a PDF, checks an analysis plan and renders a verified video. An assistant or operator supplies the analysis. This route does not require the application's API key; automatic analysis in the studio uses a configured provider.
Not yet. Interactive 3D instructions, QR delivery, a customer API and manufacturing-system integrations are on the roadmap. This website showcases the current engine and the direction of the company; it does not accept customer drawings or payments.
Explore what Assembly Animator makes possible today.