Every consumer AI app in 2026 wants you on a subscription. $9.99 a month, $19.99 a month, $29.99 a month — a recurring bill in exchange for a number of tokens you will mostly not use. Star Singer does the opposite. We sell music videos one at a time, starting at $2.99 for 15 seconds. No monthly plan, no tokens, no "credits that expire."
This was not the obvious path. Here is how we got here.
The problem with subscription pricing for creative tools
Subscription pricing works great for streaming services like Netflix or Spotify, where you consume continuously. It works terribly for creative generation tools, where usage is bursty and uneven.
If you pay $19.99 a month for an AI video tool and only make two videos, you paid $10 per video. If you make 20, you paid a dollar per video. The tool loves that variance — the light users subsidize the heavy ones. But the light users eventually cancel, because they feel ripped off, so the tool has to constantly acquire new subscribers to replace them.
We did the math. At our old $19/month plan, 62% of customers made fewer than four videos a month. Those people were paying $5 per video when a 15-second video costs us closer to $0.80 to generate. We were taking money from people who were not getting value, to fund the 10% who were making 30 videos a month.
That is a bad feeling on both sides.
Switching to pay-per-video
In January we switched. No subscription. Every video priced at what it costs to generate plus a sustainable margin: $2.99 for 15s, $4.99 for 30s, $9.99 for 60s, up to $14.99 for extended videos. Listening is free. Song generation is free (1/day).
Three things happened.
Conversion rate went up. Free users were more willing to try a single $2.99 video than to commit to $19/month.
Total revenue stayed roughly flat. We lost recurring revenue from power users, but gained from occasional users who would never have subscribed.
NPS went up 18 points. People hate subscriptions. They like buying a thing they want.
The tradeoffs
One-time pricing is harder to forecast. Your MRR charts do not exist. Cashflow is lumpier. Investors ask uncomfortable questions about "stickiness." We have had to get good at telling the story differently.
It is also harder to fund R&D. Subscription revenue lets you invest in moonshots. One-time revenue forces you to justify every feature in terms of immediate incremental sales.
We think both tradeoffs are worth it. We do not want to build a product that depends on people forgetting to cancel. We want to build one people actively choose, one video at a time.
The quiet benefit
There is also a benefit we did not anticipate. Because every video is a discrete purchase, customers treat each one as a deliberate choice. They think harder about the song, the visual style, the length. They make fewer but better videos. The content on the Star Singer feed noticeably improved in the month after we switched.
Turns out making every dollar a friction-laden decision is… actually good for the art.
What about power users
A small number of creators really do make 30+ videos a month. For them we offer optional credit packs with a volume discount — buy $50 of credits, get $60 of value. It is still pay-per-video, just cheaper per unit. About 4% of revenue now.
If you were waiting for a free trial
There is no free trial because there does not need to be one. Listening is already free. You already know if you like the catalog. The only question is whether you want to be in a video of your own. $2.99 answers that without making anyone commit to anything.
That is the whole pitch. One-time pricing, honestly priced, for a thing you either want or do not.