Making photorealistic architectural rendering fast, precise, and accessible
VizBase started from a workflow problem every architect and interior designer knows: rendering is the slowest, most expensive step in a process that lives on iteration. A single traditional render can take hours of machine time or days of turnaround from an outsourcer, so options get cut, feedback rounds stretch, and the version the client approves is rarely the version anyone had time to explore.
AI could obviously close that gap, but AI rendering arrived with a credibility problem of its own: it changes the design. Walls move, windows appear, proportions drift. VizBase was built around the opposite promise. Depth guidance, structure locks, and per-element masks hold your geometry, so what comes back is your design with photorealistic materials and light, not an AI's reinterpretation of it.
The first version was built in roughly 24 hours at a hackathon in Riga in December 2025 and launched publicly on January 29, 2026. VizBase is bootstrapped, built in Latvia, and part of the LIAA startup incubator.
What VizBase offers today
One platform that covers the working loop of a visualization project: render, specify materials, discuss, decide.
RENDER ENGINESFive engines, drafts to flagship
Photorealistic rendering with geometry preservation, from 0.5 credits per image.
MATERIALS2,000+ material library
Curated, commercially safe materials applied to any element in one click.
MATERIAL EXTRACTIONAny photo becomes a material
Click a surface in a photo to extract a seamless material with full PBR maps.
BOARDSA live canvas for decisions
Sketches, renders, and notes side by side, shared with one link.
TEAMSPACESShared studio workspaces
Projects, credits, and libraries in one place. Seats on every paid plan.
Renders typically arrive in under a minute, work with any tool that exports an image (SketchUp, Revit, Rhino, Blender, hand sketches), and there is a native SketchUp plugin from the Starter plan up.
What we hold ourselves to
The design is the contract
AI rendering earns trust by not redrawing the work. Depth guidance, structure locks, and per-element masks exist so your geometry survives the render.
The work stays yours
Full commercial rights to every render, on every plan, even after you cancel.
Your data is not training data
Uploads and renders are used to produce your results, not to train AI models. That promise is written into our terms.
The designer stays the designer
Vizbase handles the rendering grunt work. Creative direction, taste, and judgment stay with you.
Who builds it
Mikus Zarāns
TECHNICAL CO-FOUNDER & CEOA self-taught builder with a background in video production. Mikus builds the product end to end: the render pipeline, the plugin, the platform.
Līva Pērkone
BUSINESS CO-FOUNDERPreviously CEO of Helve, with experience across Nasdaq and Printify. Līva runs the business: operations, partnerships, and go-to-market.
Milestones
Our technology
VizBase combines specialized AI systems rather than betting on a single model. One system detects and masks every element in a scene; five render engines with different strengths generate the images; structural analysis extracts depth and line information so geometry stays anchored. The engines are updated as the field moves, without you changing anything in your workflow. Everything runs in the cloud and in your browser: no GPU, no installs, no render farm.
Contact
Email: support@vizbase.ai
SIA Vizbase (Vizbase Ltd.), Krišjāņa Barona iela 30A, Rīga, LV-1011, Latvia