Abstract
Hailuo AI turns a text or photo prompt into an AI-generated image or short video, sold on a credit system layered under a Play Store subscription. Recent reviews describe credits burned on generations that fail outright, subscriptions charged before a user gets a genuine free try, and charges that continue after cancellation — a billing-trust problem, not a model-quality one, on an app with 22M+ installs.
Background
For over a year Hailuo AI: Image&Video Maker has ridden the text-to-video wave to 22M+ installs on a freemium credit model: download free, spend credits per generation, subscribe for more of them.
What it does
Generates images and short videos from a text prompt or a source photo, charging a per-generation credit deducted from a paid balance or subscription allotment.
What changed
Nothing structural — but the last two months of reviews show a spike in generations that fail or come back deformed while still consuming the credit, alongside Play Billing subscriptions that start, renew, or fail to cancel without the user's clear consent.
Why it is soft
The complaints target billing integrity, not the AI itself — several of the same low-star reviewers and a handful of 5-star ones separately confirm the generations look good when they actually complete, which points at a fixable trust problem rather than a broken product.
Strengths
The backlash is about being charged for nothing, not about the app being bad at its one job — reviewers keep describing themselves as fans even mid-complaint, which is the install base a new entrant would otherwise spend years building.
Proven demand
22M+ installs and an active paying subscriber base, even amid the complaints below, show real willingness to pay for AI video generation rather than a market that has to be created from scratch.
Quality when it works
Multiple 5-star reviews single out specific generations they were happy with, asking mainly for lower prices — the underlying output is not the thing driving the 1-star pile-up.
easily my favorite app but I don't know what changed in the last few days cause there hasn't been any update, but the amount of failed generations is insane, when they do generate 9 out of 10 of them are drastically different from the prompt or deformed.
Market gap
The recurring complaints since early summer, grouped into named clusters — most-common first:
Charged without consent or after cancelling
The largest cluster by a wide margin: subscriptions that renew or charge again shortly after a purchase or a cancellation, with no clear notice, pushing users toward bank disputes and Google Play refund requests.
Credits burned on failed generations
A generation errors out or returns something unusable, and the credit is still deducted with no automatic refund or retry — the single most specific, repeated complaint in the review set.
No genuine free trial
Marketed as free to download, but the first real generation is gated behind a paid plan or a paid "trial" charge, so a user can't judge output quality before paying.
I signed up for the 3-day trial, which ends on June 28. On June 25, before the trial ended, I purchased the 1,000 + 200 bonus credits package for USD 10.43, and the credits were added successfully. Shortly afterward, I was charged another USD 11.30 without my authorization, and I received no additional credits or benefits for that charge.
I paid for a service that never worked. It is unacceptable that your app deducted my credits for failed generations caused by your own system. i wanted to create a World Cup video for my son, but unfortunately the app didn't work at all. I tried generating the video four times, and every attempt ended with a "Generation Failure" error. Despite the failures, my credits were deducted each time until I was no longer able to try again.
- Never deduct credits on a failed or errored generation. Auto-refund or auto-retry instantly instead of a silent debit — this alone answers the single loudest, most specific complaint above.
- Offer a no-card, low-res free generation before any payment. Even one or two free tries removes the false-advertising accusation and lets output quality speak before asking for a card.
- Make cancellation instant and self-service inside the app, with billing dates shown in-app. Every charged-after-cancelling report above traces back to a cancel flow users couldn't find or trust — put one in the app and confirm it on-screen.
Build complexity
Estimated, not derived: one developer, part-time, no funding and no paid acquisition. The generation itself is bought from an existing text-to-image/video API, so the build is mostly billing, trust, and polish, not model research.
| Workstream | Weeks |
|---|---|
| Generation API integration (image + short video) | 1.5 |
| Credit ledger + Play Billing, refund-on-failure logic | 2.0 |
| Auth + gallery/history | 1.0 |
| Free tier + abuse/rate-limit guardrails | 1.5 |
| Store listing, subscription disclosure, polish | 2.0 |
| Total | 8.0 |
Expected revenue
Assumes a $7.99/month subscription undercutting the pricing complaints above, reached without paid acquisition — word of mouth and store search inside a category already searched by millions.
| Month 12, monthly | ||
|---|---|---|
| Installs | 40,000 | |
| Activated | 12,000 | 30% |
| Retained 30d | 3,600 | 30% |
| Paid | 750 | 21% |
| Revenue | $6,000 |
Biggest challenges
Not the AI integration — the business is entirely dependent on a third-party generation API's per-call cost, and re-earning store trust while directly inviting comparison to an app whose own reviews call it a scam.
Margin tied to a vendor's pricing
As a thin wrapper on someone else's model, a per-generation price change upstream can erase margin overnight — unlike a product that owns its own compute.
Store trust from zero
Launching at 0 reviews directly against an incumbent flagged "scam" in its own reviews means one early unauthorized-charge complaint can sink a young app's rating far faster than it dents a 22M-install one.
AI angle
The product is already built on AI end to end — the genuine angle is using a second, cheap AI call to police the first one, turning the loudest complaint above into an automated fix.
Without AI
A human support team manually reading "my video didn't render" tickets and issuing refunds by hand — slow, expensive for a solo builder, and exactly what the reviews above say the incumbent doesn't do.
With AI
A cheap vision-language model call checks the finished generation against the prompt before the credit is finalized, auto-refunding or auto-retrying a failed or mismatched result with no support ticket — days of integration, not weeks, and it folds into the estimate above.
Conclusion
Build it. 22M installs and a wall of near-identical billing complaints sitting on an app whose AI output quality isn't seriously in question — the whole opportunity is trust, not technology.
8 weeks
Time to v1
$7.99/mo
Credits with auto-refund on failure
$6,000
Revenue at month 12
Hundreds
Of near-identical unauthorized-charge complaints