mnml

mnml.ai is an AI rendering studio aimed squarely at architects: you feed it a sketch, a site photo, or a viewport export from SketchUp, Revit, Rhino or Blender, tell it what the image is (exterior, interior, masterplan, landscape, plan or product), and it produces a presentation-grade render. What separates it from general-purpose image generators is that the whole workspace is built around architectural design development — iterations, comparisons, client review, region-level refinement — rather than around a prompt box. The vendor claims more than 2.5 million registered architects and designers and use by teams at 2,400 firms and 250 universities, naming Gensler, SOM, HOK and OMA among them; those are vendor claims, though the plugin ecosystem behind them is real and freely downloadable.

What it actually does

The product has two working modes that share one project library and one credit balance. Studio takes a single image all the way: import, pick the image type and a style (photoreal, CGI, sketch, model, watercolor), render, then refine one region at a time until the image is client-ready, and optionally turn it into a short clip. Flow is the comparison mode: an infinite board where renders land beside their source, up to four variations at a time, each render recording its own lineage — sources, expert, prompt — so a failed variant can be re-run from its recorded submit and a good one branched without losing the others. Flow also supports sticky notes, group frames, and live cursors for review sessions with clients or teammates.

Around those two modes sits the one-shot toolset: exterior AI, interior AI, virtual staging, home design, masterplan and landscape tools, a design assistant that changes materials and lighting from text instructions, style transfer from a reference image, and an enhancer/upscaler that goes to 8K. Video generation (short clips) is included with monthly allowances per plan. There are also free plugins — this is the part competitors often lack — that render the view you are modeling in without exporting: mnml for SketchUp (SketchUp 2026, Windows and Mac), mnml for Revit (Revit 2024–2027, Windows), and mnml for Rhino 8 with a Grasshopper tab (Windows and Mac). The Grasshopper integration batches up to eight design options behind one confirmation, and re-solving a Grasshopper definition does not spend credits. The site also advertises mnml inside ChatGPT and Claude.

Credit economics: credits are the currency, and each engine costs a fixed amount shown on the button before you click — main tools commonly 10 credits per design, auxiliary tools 1–5, the video workflow around 100. At time of writing the pricing page lists Lite at $29/month for 1,000 credits (roughly 100 Fast renders or 30 Ultra renders, two concurrent renders, 10 video generations), Plus at $49/month for 3,000 credits (four concurrent, 30 videos), and Studio at $119/month for 10,000 credits (six concurrent, 100 videos). All plans include all tools, 8K upscaling, PNG and DXF export, a commercial license, and credit rollover. Every new account gets free credits that do not expire, with no card required. A one-time credit pack (10,000 credits) is sold for projects without a subscription.

My testing notes

First, the input flexibility is the product's real strength. In my testing, feeding it a rough SketchUp viewport and a hand sketch of the same building produced coherent, geometry-respecting renders from both — the massing stayed true to the input rather than drifting into generic "nice house" shapes the way general image models do. The image-type selector (exterior vs. interior vs. masterplan) makes a visible difference to the output: it behaves like a routing layer to different engines, not a style label.

Second, the plugin workflow is where the time saving shows up. Rendering the view I was modeling in — without exporting, naming, uploading, re-prompting — collapsed the loop between "adjust the massing" and "see the render" to about a minute. The price-on-the-button disclosure is a small design choice I found genuinely useful: I always knew what a click would cost before spending, which is not true of most credit-based creative tools. One caution from a third-party teardown I cross-checked: the Revit add-in, at the version tested by archigenai, captures the viewport but does not read BIM material assignments — it treats the Revit view essentially like a screenshot, so if you want material fidelity from Revit, an Enscape render as the input image produces better results than the direct plugin path. My own experience with viewport-style inputs aligns with that: the tool is only as good as the lighting and material information visible in the input image.

Third, Flow boards are the feature I did not expect to like. Comparing four variations next to the source — and being able to branch the one that worked while keeping the rejects visible — matched how design reviews actually happen. The lineage record (what produced what) is the kind of thing that matters two weeks later when a client asks "can we go back to the other sky?"

Where it stops

• No CAD/BIM translation. mnml renders images of your model; it does not read your model's data. Revit material graphs, schedules and constraints are outside its reach (see the archigenai finding above).
• Batch rendering is limited. Grasshopper batches eight options; general batch processing across many images is still described as roadmap in third-party testing.
• Credit economics need planning. Video at ~100 credits per run and Ultra renders at triple-digit credit costs can drain a Lite plan quickly; heavy users report budgeting carefully (Tooliverse flags the same).
• Purchasing is generally final per third-party summaries of the terms, and plan structures have shifted repeatedly — several directories list tiers that no longer exist.
• Loose input sketches degrade output. Messy, ambiguous sketches produce unreliable geometry; the tool works best from clean linework or model views.
• It is not a rendering engine replacement. For final-frame V-Ray/Enscape quality with true materials and lighting, this is a concept and iteration tool, not the last step.

Verdict

For architects who need client-facing visuals during design development — concept, SD, DD, competition boards — mnml.ai is one of the more complete offerings I have tested: real CAD plugins, a comparison board with lineage, 8K output, and per-click cost disclosure. The gaps are what you would expect from the category: no BIM intelligence behind the images, credit costs that need managing, and a Revit integration that is shallower than the homepage implies. Weighing depth of core functionality at 40%, value at 20%, usability at 20%, support and ecosystem at 10% and reliability signals at 10%, I rate mnml.ai 3.7. It earns that on breadth of the toolset and the plugin loop; it does not go higher because the output ceiling is your input image's quality and the BIM story is marketing rather than mechanics. For a solo architect or small studio visualizing options fast, the Plus plan is the sweet spot the numbers support.

One thing I want to add on positioning, because it shapes whether this tool fits your pipeline. I treat mnml.ai as a design-development instrument rather than a delivery renderer. In my testing the outputs were convincing enough for client conversations at concept and schematic stages — a board of four exterior options rendered from a massing study will carry a review meeting. But when I compared the same input against a full Enscape or V-Ray frame, the differences in material response, glass behavior and artificial lighting were exactly where you would expect a diffusion-style engine to approximate rather than compute. That is the honest ceiling: the tool compresses days of visualization work into minutes for early phases, and it does not retire your rendering engine for the final package. Plan your credits around that division of labor — many Light renders during design, fewer Ultra renders for the ones that survive to the presentation.

Sources

• https://mnml.ai/
• https://mnml.ai/pricing
• https://mnml.ai/plugins
• https://mnml.ai/sketchup
• https://mnml.ai/revit
• https://mnml.ai/rhino
• https://mnml.ai/grasshopper
• https://archigenai.com/mnml-ai-review-architects-2026.html
• https://therundown.ai/tools/mnmlai
• https://itechguides.com/products/mnml-ai
• https://tooliverse.ai/tools/mnml-ai
• https://recatools.com/ai-directory/mnml-ai