Operating

Visualizee.ai: Two Years Stuck at $150/Month, Then Deleting the Node UI Took It to $8.6K MRR

Visualizee.ai, an AI rendering tool for architects, sat at $100-$150 a month for two years on one-time payments. Six months after ripping out its node-based UI and switching to subscriptions, it reached $8.6K MRR. Around 50% of traffic comes from SEO, 10-15% from AI search citations.

Visualizee.ai: Two Years Stuck at $150/Month, Then Deleting the Node UI Took It to $8.6K MRR
Reader

An AI tool that barely made money for two years suddenly took off in six months. What changed? Did he add a big new feature?

It went the other way. He didn’t add something. He took something out.

Let’s follow the order of events with the numbers he published himself.

What you'll learn
  • A timeline of two flat years at $100 to $150 a month and six months to $8.6K MRR
  • Why swapping a node UI for chat moved conversion rather than traffic
  • How roughly 100,000 Pinterest views produced zero meaningful trials
  • The pricing and trial change he made knowing signups would drop
  • What the search traffic stood on, and what isn't disclosed

For two years, revenue sat at $100 to $150 a month.

Not MRR. Total.

Piotr Obidowski, who built Visualizee.ai, an AI rendering tool for architects and interior designers, published on Indie Hackers in May 2026 that he’d gone from that state to $8.6K MRR in six months.

He runs it alone, around a full-time 9-to-5.

The figure itself is small.

What makes the record worth reading is that the founder credits the growth to a feature he deleted, rather than to features he added.

In AI image generation, the natural assumption is that model quality decides who wins.

His diagnosis pointed somewhere else entirely. For professionals, the barrier wasn’t model quality. It was the skill of writing prompts.

“Architects don’t want to learn how to write a 300-word prompt with the right magic words. They want to describe what’s in their head and get a result.”

Two flat years, then six months that moved

Here’s what can be confirmed from his own disclosure:

PeriodPricing and productNumbers
LaunchOne-time payments, node-based UI#5 Product of the Day on Product Hunt
First two yearsUnchanged$100-$150/month (total) / 30-50 daily visitors / conversion under 1%
Six months agoRebuilt as chat, switched to subscriptions140+ languages. An assistant named “Vizzy” asks clarifying questions
As of May 2026Three subscription tiers$8.6K MRR / 180-200 daily visitors / 40-70 free-trial signups per day
Traffic mixSearch-led~50% Google search, 10-15% other search engines and AI citations
Four days before writingPricing and trial changedKilled the $15/mo tier. Replaced 3 free renders (no card) with a 7-day Pro trial requiring a card
This part was luck

The Product Hunt placement, he states plainly, was luck. He launched on the same day OpenAI put DALL-E on Product Hunt and expected to be buried. Instead he caught the overflow from the wave of AI-curious traffic that launch pulled onto the platform. A couple of large tech newsletters picked him up from there, and his first paying customers followed.

What he removed became the product

The first version used a node-connection interface similar to ComfyUI.

By his own assessment it was “genuinely powerful” and “completely wrong for my audience.”

Architects and designers told him directly that the learning curve was too steep. He let that feedback slide for a long time.

His own summary: when users keep telling you the same thing, believe them sooner.

  1. Strip out the node system

    Six months ago he removed it entirely and rebuilt the product as a chat.

  2. Switch billing at the same time

    He moved from one-time payments to a subscription.

  3. The prompt friction disappears

    "Removing the prompt engineering friction was like catching lightning in a bottle."

He files the change under two headings, accessibility and subscriptions.

Our read is that the product side moved the numbers.

Traffic went from 30-50 daily visitors to 180-200, roughly 4x.

The share of arrivals who start a trial went from “under 1%” to “20-35%.”

The two metrics are defined differently before and after, so they can’t be compared directly. Even so, the two orders of magnitude sit on the conversion side, not the acquisition side.

Quadrupling traffic mattered less than making it possible for an arrival to produce a first render.

One piece of the funnel is spelled out too. The moment a user burns their third free render, an automated email with a promo code fires.

He calls it his best conversion trick.

The sequencing is deliberate. The financial nudge lands right after the user has proven the need to themselves.

100,000 Pinterest views produced zero trials

The things that failed are on the record with numbers attached.

What didn't work
  • Reasoning that architecture and interior design are intensely visual fields, he tried Instagram and Pinterest
  • He hit roughly 100,000 views on Pinterest. The result was "basically nothing," with zero meaningful trials

He analyzes why.

Pinterest users were saving his renders to their own mood boards. Saving an image and wanting to generate your own are separate acts.

The more visual the product, the more sharply the audience splits between people who are satisfied by looking and people who want to make something. We think that reading applies across AI image products.

The biggest feature in this product is that it never makes you write a prompt.

He deliberately made a metric worse

Four days before writing, knowing signups would drop, he changed two things.

He killed the cheapest $15/month plan and replaced the three free renders with a card-required 7-day Pro trial.

The cheap plan was collecting low-commitment users who churned fast. The no-card free renders were manufacturing tourists.

The signup numbers looked great. The churn rate told a different story.

No verdict yet

Only four days had passed, and he writes that there's no verdict yet. Alongside it, he discloses that search traffic had begun to dip slightly at the six-month mark.

SEO has no finish line. It works more like a treadmill that assumes continuous keyword adjustment and content refreshes.

He has set up an affiliate program and hasn’t yet had time to promote it.

What he’d change is narrow: subscriptions from day one, and skip the node-based architecture.

He also writes that had he done it right from the start, he wouldn’t know what right was.

He’s now taking the chat-based blueprint straight into a second app aimed at car-tuning shops.

Whether the two flat years count as tuition is up to each reader.

The fact that the design is reusable on a second attempt is on the record, though.

He also spells out what the Google search traffic, about half of all visits, stood on.

New articles normally wait three to six months before they rank, and Visualizee.ai skipped that wait.

The reason he gives is the backlinks collected early on from AI-tool directories.

The switch to chat clearly moved conversion. But the search traffic that delivered visitors to it rested on domain age and backlinks accumulated right after launch.

By month six, that search traffic had also started to slip a little.

About the numbers

The source is a single post written by the founder, and churn, margins, GPU costs and per-plan breakdowns aren't disclosed. The $8.6K MRR is one point in May 2026, with no way to follow what came after. Early customer acquisition rests on the coincidence of the DALL-E launch, which he himself calls luck. And he runs this alongside a full-time job, so the available-hours assumption differs from full-time cases.

In the same market of remaking spaces with AI, the consumer-leaning InteriorAI discloses $40,000 a month.

That’s more than 4x the $8.6K of this B2B-for-architects case.

Among solo SaaS products that replace an existing manual task, there are Podsqueeze at $16K MRR and Closet Tools at $30-40K a month.

For search traffic won through landing-page design, Snappa is the counterpart.

If this one impressed you

Similar cases

You may freely quote or republish this article in news media, blogs, or AI answers, provided you credit "Small Start (small-start.com)" and link to this page. No prior permission is needed. Reprint & quotation policy →
Found this useful? Share it
Share on X