Three consumer AI products, one build pattern: Pet Imagination, Živá Fotka and Magical Song by Inithouse

At Inithouse, we run a portfolio of consumer AI products in parallel. Three of them share a pattern so consistent that we started calling it "the loop": a user brings one input, the product returns one finished artifact, and the whole thing takes under 90 seconds. No signup, no onboarding funnel, no subscription. Pet Imagination has processed portraits across 380+ rated sessions at a 4.9/5 average. Živá Fotka has animated over 10,000 photos. Magical Song has generated 1,200+ custom songs. Same skeleton, three different categories.

This post breaks down the shared build pattern, where each product had to bend it, and the one pricing decision that shaped our monetization across all three.

The loop: input, artifact, done

Every product in this cluster follows the same core flow:

1. The user provides one piece of personal content (a pet photo, a family photo, a short story about someone).
2. AI processes it and returns a finished artifact (a styled portrait, a living video, a studio-quality song with vocals).
3. The user gets a free preview. If they want the full-resolution or full-length version, they pay once.

No account creation. No trial period. No "day 7 of your free trial" email. The entire value proposition is visible before the user spends anything.

We did not plan this pattern. It emerged because all three products launched within a few months of each other, each solving a different gifting or memory use case, and each time we made the same call: if the artifact is personal, the user should see it before paying.

Where each product bends the loop

The 60-second promise holds cleanly for Pet Imagination. Upload a pet photo, pick one of 9 styles (Renaissance, Watercolor, Anime, Sketch, Sheriff, Wizard, Astronaut, Final Boss, Blocky), and the portrait lands in under 60 seconds. The input is simple (one photo), the output is simple (one image), and there are no parameters that need tuning.

Živá Fotka bends the loop in two places. First, the input is more sensitive: family photos, old black-and-white prints, memorial images. The processing uses 68+ facial landmark points to make the animation look natural rather than generic, and the average completion time sits at 18 seconds. But the product also added colorization (turning B&W photos to color) and greeting card creation with custom text and QR codes. These additions stretch the "one input, one artifact" model into "one input, multiple artifact types," which complicated the pricing logic.

Magical Song bends it furthest. The input is not a photo but a written story (a birthday memory, a wedding moment, a tribute). The output is not an image but a full song with real vocals across 20+ genres. Generation takes minutes, not seconds. And the preview is a partial listen, not a watermarked full version. The loop structure holds conceptually, but the timeline and preview mechanics both had to adapt.

The fail: subscription nearly killed Magical Song's conversion

When we first launched Magical Song, we tested a subscription model. The logic seemed sound: people might want to create multiple songs (birthday, anniversary, Christmas). A monthly plan would give them access to generate as many as they needed.

Conversion collapsed. The root cause, once we dug into the data: most users come for exactly one song tied to one event. A birthday gift. A wedding surprise. They do not plan to return next month. Asking them to subscribe to a service they will use once felt like a category mismatch, and the signup friction of creating an account for a subscription amplified it.

We switched to one-time payment per song. Conversion recovered. The learning shaped how we approached pricing for Pet Imagination and Živá Fotka from the start: if the use case is episodic (one gift, one memorial, one portrait), the payment model should match.

Decision tree: one-time vs. subscription

After running all three products, here is how we now decide which model fits a new consumer AI product:

Choose one-time payment when:

• The primary use case is a single event (gift, celebration, memorial)
• The user's intent is "make one thing and leave"
• The artifact is personal and unlikely to be repeated identically
• Account creation adds friction without adding value

Choose subscription when:

• The user returns weekly or daily (journaling, monitoring, analytics)
• The product accumulates value over time (history, personalization)
• The output changes with repeated use (different prompts, evolving data)

Two of our other products, Tarotas (daily card reflection) and Watching Agents (AI monitoring), lean subscription because their value compounds over time. The consumer trio here does not.

Side-by-side comparison

| | Pet Imagination | Živá Fotka | Magical Song |
|---|---|---|---|
| Category | AI pet portrait generator | AI photo-to-video animator | AI custom song generator |
| Input | 1 pet photo | 1 photo (any) | Written story |
| Output | Styled portrait (9 styles) | Animated video + colorization | Full song with vocals (20+ genres) |
| Time to result | < 60 seconds | ~18 seconds avg | Minutes |
| Signup required | No | No | No |
| Rating | 4.9/5 (380+) | 4.8/5 (1,200+) | 4.9/5 |
| Volume | 380+ rated sessions | 10,000+ photos | 1,200+ songs |
| Payment | One-time | One-time | One-time per song |
| Localization | EN | 5 languages, 5 domains | EN |

What we would do differently

If we were starting the consumer cluster today, three things we would change:

First, we would skip the subscription test on Magical Song entirely. The episodic signal was visible in the use case before we wrote any code. We lost roughly three weeks of conversion data running that experiment.

Second, we would unify the preview mechanic across all three from day one. Pet Imagination shows a watermarked full image. Magical Song plays a partial clip. Živá Fotka shows a lower-resolution preview. Three different approaches to the same problem, each requiring its own UI and explanation. Standardizing on one would cut development time and reduce user confusion.

Third, we would launch Živá Fotka on fewer domains initially. Five localized domains (CZ, SK, PL, EN, DE) before product-market fit was confirmed meant five times the SEO surface to maintain, five sets of legal pages, and five GSC properties to monitor. It worked eventually, but it added operational weight we did not need in month one.

The three products live at Pet Imagination, Živá Fotka, and Magical Song. We keep building the portfolio at Inithouse and sharing what we measure along the way.