Abstract
Hula turns selfies into AI avatars and stylized videos, but reviewers overwhelmingly report paying a subscription and then hitting a second "coins" paywall before generating a single result โ a billing-trust failure layered on top of a real generative-AI cost problem.
Background
Reviewers describe Hula - AI Video Generator as effectively a wrapper around third-party generative video/image models.
What it does
Turns a selfie into AI-generated avatars, stylized videos, photo-to-video effects, and sticker packs using generative video/image models.
What changed
Nothing about the AI itself โ the complaint is entirely the pricing model: users pay a weekly or yearly subscription, then discover they must separately buy "coins" to generate anything at all.
Why it is soft
This is a billing-trust failure, not a product-quality one โ reviewers who never even got to see a result still call it a scam, which is the cheapest kind of gap to close.
Strengths
When it actually generates something, several reviewers call the output amazing and worth it; the core idea โ instant stylized AI selfies, avatars, and stickers โ clearly has real appetite, evidenced by installs despite the near-uniformly hostile pricing reviews.
Market gap
The recurring complaints, most-common first:
Subscription plus separate coins
By far the most common complaint: a paid subscription that still requires buying "coins" separately just to use a feature already paid for, described repeatedly as a scam.
No free preview
No way to see a single free generation before being asked to pay โ reviewers say competitors let them try before buying.
Hard to cancel, no refunds
Difficult-to-find cancellation and no refunds, even within hours of a mistaken purchase.
Inconsistent output quality
On the rare occasions people got past the paywall, results sometimes swapped faces or didn't resemble the uploaded photo.
I bought yearly service and every week I was limited but I could still use it which was fine but then they changed their model with no regard to anyone who purchased the yearly and now you need to pay with coins
I paid for a subscription thinking everything would be included, but then I found out I still have to buy coins just to actually use the features. That feels like a complete rip off.
- One price, no separate coins. Whatever the subscription costs, it should include a fixed, clearly-stated number of generations with nothing else to buy.
- Let new users generate one free result before paying. The single most requested fix, and the cheapest one to ship.
- Make cancellation and refunds self-service in-app. Not a support ticket black hole โ that alone would flip a large share of the one-star "scam" reviews.
Build complexity
Estimated for one developer, part-time, no funding, no paid acquisition, using an off-the-shelf third-party AI image/video generation API rather than training a model.
| Workstream | Weeks |
|---|---|
| Selfie upload + generation pipeline (third-party API) | 2.0 |
| Style/template library + gallery UI | 1.5 |
| Single transparent subscription billing | 1.0 |
| Sharing/export + polish | 1.5 |
| Total | 6.0 |
Expected revenue
Assumes a flat $9.99/month subscription that includes a fixed number of generations, with API costs absorbed into that price rather than billed separately.
| Month 12, monthly | ||
|---|---|---|
| Installs | 8,000 | |
| Activated | 3,400 | 43% |
| Still on at 30d | 1,500 | 19% |
| Paid | 300 | 4% |
| Revenue | $3,000 |
Biggest challenges
Video-generation API costs are the real constraint โ a solo builder pays retail rates to third-party generative-video providers per generation, so the "coins" model competitors use exists partly for real cost reasons; pricing has to cover API cost per generation or the business loses money on its most active users.
AI angle
This app's entire product is the AI angle โ the opportunity isn't a cheaper way to build the generation pipeline, it's an honest wrapper around the same generative-video APIs incumbents already use, competing on pricing transparency rather than technology.
Conclusion
Build it. A real, well-documented billing-trust complaint and genuine demand for the concept, tempered by thin margins once real generative-AI API costs are priced in.
6 weeks
Time to v1
$9.99/mo flat, no coins
Price and model
$3,000
Revenue at month 12
Reviewers calling it a scam without ever seeing one result
The fact that decided it