Manual.

Getting started.

A project in vizt holds one site and every model you propose for it. Six steps take it from an export to a link your client can walk.

1. Create the project. Name it after the site or the client — you can rename it later from the project rail.
2. Upload a model. Anything in the list below converts automatically. Each upload is a separate version, so an alternative scheme sits beside the first rather than replacing it.
3. Place it on the site. Drag the footprint onto the satellite image and rotate until the walls align. This is what makes sun, compass and on-site GPS true.
4. Check the layers. vizt reads your material and layer names to decide what each object is and how it renders. Override anything it got wrong — no re-export needed.
5. Create a client link. One link per client, optionally with a password and an expiry date. Revoke any of them without touching the others.
6. Read what comes back. Notes your client pins in the viewer arrive on the project page, each tied to the spot and the viewpoint they were standing at.

Exporting models.

vizt accepts GLB (best), OBJ, FBX, DAE, STL, PLY and IFC, up to 300 MB. Everything is converted and compressed on upload.

Three rules apply whatever you model in: work in metres, hide reference and construction layers before exporting, and ignore orientation entirely — you place and rotate the model on the map inside vizt, so the export does not need to face anywhere in particular.

Rhino

1. Check units: Tools → Options → Units → Model = metres.
2. Hide or delete construction layers, then select everything to publish.
3. Rhino 8: File → Export Selected → glTF Binary (*.glb). Keep materials checked.
4. Rhino 6/7 (no glTF): File → Export Selected → OBJ. In meshing options choose Smooth & slower, tick Weld vertices and Export material definitions.

Revit

1. File → Export → IFC.
2. Setup: IFC4 Reference View — IFC 2x3 also works.
3. Export only the 3D view you want published; hide annotation categories first.
4. Walls, roofs and openings arrive with their true geometry.

Archicad

1. File → Save As → IFC.
2. Translator: General Translator (IFC4) — geometry only is enough.
3. Filter to the published story range and layers before saving.

SketchUp

1. File → Export → 3D Model → COLLADA (.dae) — available in Free and Pro.
2. Tick Triangulate all faces and Export texture maps.
3. Pro users can export OBJ or FBX instead; both upload fine.

Blender

1. File → Export → glTF 2.0 (.glb).
2. Tick Apply Modifiers. Leave +Y up as it is — handled automatically.
3. Compression is optional; vizt re-compresses either way.

Site placement.

Placement is what turns a model into a building on a plot. Until it is done, the sun has no bearing, the compass has nothing to point at and On Site cannot find you.

Open Site & calibration on the project page. You get the satellite image of the plot with your model's footprint over it. Drag the footprint into position and set rotation until the walls line up with what you can see on the ground. Adjust north if the survey and the satellite disagree.

The cyan pin is where a client's visit begins. Drop it at the entrance, or at whichever view you want them to arrive on — it is the first thing they will see, and it is worth choosing rather than accepting.

One site, several proposals

Placement belongs to the project, so every model in it sits on the same plot by default — which is what makes two schemes genuinely comparable. When one proposal needs to sit differently (a shifted footprint, a different orientation), switch that model to its own calibration on the same page. The plot itself — latitude, longitude, time zone — always stays with the project.

Layers & materials.

The viewer understands your model through layer and material names. For each object vizt reads the material name first; if the material is unnamed — a Rhino "default" material — it falls back to the nearest named parent layer or node.

Matching is case- and accent-insensitive (Cubierta = cubierta), the keyword only has to appear somewhere in the name (3.4.- TEJAS matches teja), and Spanish and English keywords both work.

Each name drives two independent decisions: the category — what the object is, which controls viewer behaviour — and the material — how it renders. puerta vidrio is a Door you can walk through, rendered as glass. Both are overridable per layer from the project page, with no re-export.

1 · Categories — behaviour

Checked in this order; first hit wins, so puerta muro is a Door, not a Wall.

2 · Materials — appearance

Also first-match-wins, in exactly this order.

3 · Walking and collision

With Solid walls on — project page, Layers & materials — visitors collide with every object except Door layers.

• Put door leaves on their own puerta / door layer, so a closed door never traps a visitor in a room.
• Door frames can stay on the door layer. Glass panes need their own vidrio layer if you want them to stay solid.
• Openings without a door object need nothing — collision only blocks where geometry actually is.
• Naming walls is optional for collision, since everything solid blocks. It documents intent, and lets you flip a mislabelled layer to Door in one click.

4 · Gotchas

• First match wins: viga metal renders as metal, not wood. suelo hormigón renders as concrete — name it just suelo for the floor finish.
• One layer per real-world material is the golden rule. Split mixed layers at the source.
• A layer only appears in the list when its objects carry a named material, or — if the material is unnamed — when a parent node carries the layer's name. Everything else shares the (default) row.
• Category and material are independent: a cubierta vidrio skylight hides with the Roofs toggle and renders as glass.
• Keywords match anywhere in a name, so muros_bajos is a Wall. Check the "Auto (…)" hints on the project page to see what was actually detected.

Viewer navigation.

The viewer opens in third person, at eye height, at the point you dropped the cyan pin. The avatar is there to give scale — without a body in the frame, a room has no size.

Moving

• Walk the design — the default. Drag the wheel to walk; the camera follows behind. On a keyboard, the arrow keys and WASD do the same.
• Fly — leaves the ground for an overview or a look at the roof.
• Top — straight down, for reading the plan and the relationship to the plot.
• Double-click or double-tap anywhere — the avatar walks to that spot. Works in every mode, and is usually faster than steering there.
• Speed — folded behind a tab at the top of the screen. C slower, V faster.

Light

The Sun tab sets the hour directly; day and month live in the menu. Shadows are computed from the real position of the sun at your site's latitude and longitude on that date, which is why placement has to be right before a sun study means anything.

Seeing in

• Roofs and Terrain hide those layers so you can look into the plan from above.
• Layers exposes any layer you published as client-togglable.
• Measure takes a distance between two points on the model.
• Notes pins a comment where the client is standing, with their viewpoint saved alongside it.
• Views jumps to the viewpoints you saved for them.

On Site.

On Site puts the proposal on the real plot, seen through the phone, standing where the building will be. It is the one feature that needs the client to be somewhere specific.

• Phones and tablets only. It needs GPS and a compass, so it does not appear on desktop.
• Needs HTTPS — which every vizt link is. Browsers refuse location on plain HTTP.
• The browser asks permission for location and motion. On iOS, motion has its own prompt; if it is declined the view still works, following drag instead of the phone's heading.
• vizt waits a few seconds for the GPS to settle before anchoring the model, because the first fix is usually the worst. Place me now skips the wait.
• The model is anchored using the placement you set on the map, so accuracy on site is the accuracy of your calibration. If it sits a metre off, nudge the heading trim rather than re-walking the plot.

Troubleshooting.

Still stuck? Send us the file and what you expected to see: hello@monomedialab.com.