The monthly fee is the number on the billboard. If you make pictures often, it can turn out to be the smaller part of what you spend.
Nothing sneaky is going on. Companies are simply passing on a genuine difference in what things cost them to run. It only becomes easy to manage when you do one bit of maths before you sign up, not after your card has been hit.
What the meter is really measuring
Sending you a text reply costs the operator a tiny fraction of a cent. Producing a picture costs noticeably more, since it ties up a GPU for every single image. Voice is charged by the second of audio, and video costs far more than either.
That explains the pricing pattern you see everywhere: chat that's generous or "unlimited", a cap on images, voice sold in minutes, and video reserved for the top tiers. The chat and the pictures come from separate engines with very different running costs, as how the media side works lays out.
So read "unlimited" on a pricing page as "unlimited text". It's honest about the cheap thing.
Three ways apps bill for pictures
Bundled with a cap. A daily or monthly quota comes with the subscription. This is the easiest to plan around, because the most you'll ever pay is the plan price.
Separate credits. The plan pays for conversation, and pictures come out of a balance you keep topping up. Nothing limits the worst case, and that's the whole trouble.
Stepped quotas. Pay more, get more images. Fair enough, provided you price the tier you'll genuinely end up on and not the cheap entry one.
The first model is why Candy AI sits on top of our ranking at US$12.99. With chat, images and voice all in one subscription, you never have to make this call. That's an advantage in how it's priced, not in what it can do.
Do this sum before paying
Two numbers, about five minutes.
First: how many pictures a week will you honestly want? Aim a little high. In the first month, when everything is new, most people guess about half the real figure.
Second: what would that cost under each model?
- All-in plan: the subscription price, done.
- Credits: pictures per week, times 4.3, times the per-picture price of the bundle you'd realistically buy, plus the subscription.
When the credits total beats the all-in total, which for more than a handful of pictures weekly it usually does, the dearer-looking plan is the cheaper one. It's the classic slip in this category: picking the low sticker price and paying more overall.
How credit packs are priced
The pattern never changes. A small pack has a steep price per credit, and a big pack has a far lower one. That slope is deliberate, because it nudges you up the ladder. The larger pack really is better value per picture.
Two questions almost never answered on the checkout page:
Do credits expire? Frequently. A big pack bought for the discount stops being a bargain if half of it lapses.
Do failed attempts cost credits? A refused prompt, a filter block, or a picture you bin straight away can each still use one in many apps. Spend a month trying out prompts and that adds up to real money. It's also the number one gripe about credit systems.
The month-one skew
You'll make far more pictures in month one than in month four, because the novelty is real and it wears off.
That cuts two ways. Don't buy a huge credit pack in week one based on week-one habits; you'd be planning around your peak. And don't lock into an annual plan built around heavy media use on that same evidence. Monthly, annual or credits walks through the trade-off.
Where your money goes
It helps to know why the caps exist. You're paying for GPU time, and the GPU bills the operator whether or not you like the result. An app that handed out endless pictures would be covering the gap out of subscriptions, and that arrangement doesn't last.
Apps doing more with media, like Secret Desires, which makes images and short video and scores 4.3 in our ranking, run at a higher cost base. How freely an app hands out media is a cost decision, and you'll see it in the plan structure long before the marketing admits it.
In short
Work out the pictures you want each week, price both models, and favour the one where the worst case is a figure you already know. Then check for expiry and for charges on failures.
Those five minutes beat any comparison table, ours included.
