getimg.ai against a receipt I already have

I don’t have to guess what bulk image generation costs, because I wrote the bill down. Last month I generated 402 images for a single clip-art book, on Replicate, from a script. The total was $1.84. The human reject rate was 3.5%. The API failure rate was zero. All of that is itemised by model and by run in the build log for that book, because the only honest way to write about image pricing is to show a receipt.

So when getimg.ai accepted me into its affiliate program this morning, I did the thing I’d want someone to do for me. I put its published prices next to my own measured ones and worked out where a hosted studio actually earns the difference — and where it very clearly doesn’t.

Let me state one thing plainly before any number appears: I have not used getimg.ai. What I have done is push hundreds of generations through a metered API for a product I sell, which means I know this category from the inside even though I don’t know this product from the inside. Those are two different kinds of knowledge and I’m not going to blur them for the sake of a commission.

I already measured what bulk image generation costs me. The only question left is where a hosted studio earns the gap.

What getimg.ai actually is

getimg.ai is a hosted generation studio: one browser interface on top of a rotating roster of other companies’ models, plus editing tools around them — create image, create video, upscale, resize, remove background, create music, create speech. It is a workbench, not a model.

The model roster is the part that dates fastest. As of today the front page names GPT Image 2.5 (Sunburst and Flare), Qwen Image 3.0 Pro, Grok Imagine Image 2.0, Grok Imagine Video 1.5, and Wan 3.0. The developer catalogue adds Seedream, Nano Banana, Z-Image Turbo and FLUX. That is a genuinely current lineup, and it is the strongest thing about the product: you are not locked to one house model, and you do not maintain the plumbing when a new one lands.

There is a second product hiding inside the first, and most reviews blur them. The studio subscription and the API are separate businesses with separate bills. getimg.ai’s own FAQ is blunt about it: “The API credits are not included in the subscription plans. Subscription is not required to use the API.” If you are the kind of person who generates images from a script, the subscription table below is largely irrelevant to you, and almost every review I read got that wrong.

The subscription pricing, plainly

Four tiers, credit-based, monthly or yearly at roughly a 20% discount. These are the numbers on getimg.ai’s own pricing page today, with the image counts it publishes for each tier at 1K resolution:

The effective per-image column is mine, from dividing the monthly price by the published 1K image count. Two things the page states that matter more than the headline price: credits do not roll over to the next month, and 2K and 4K output costs roughly double and triple, so Entry’s sixty images become thirty or twenty. Video and upscaling burn credits much faster still.

And a conflict I have to flag rather than resolve. The current pricing page shows no free tier, and independent coverage from earlier this year says the free plan was discontinued. Yet several review sites still describe “100 free credits a month.” I could not find a free plan on the live pricing page today. Treat the free tier as gone until you see it yourself.

The API is a different product, and it is priced honestly

This is the part I found genuinely good. The API is pay-as-you-go, per image, no subscription and no minimum spend, and getimg.ai publishes a per-model rate card. Its stated range is $0.015 to $0.18 per image, with video from $0.022 per second. A sample from the published catalogue:

Every response carries a usage object stating what the request cost. After a year of credit systems that hide the meter, I’ll take that seriously.

Now put it against my receipt

My run was 402 generations for $1.84. The bulk of it — 370 images — went through flux-schnell on Replicate at $3.00 per thousand output images, or $0.003 each. The test batches and reject re-runs went through flux-dev at $0.025 each. I re-checked both rates on Replicate’s pricing page today and they are unchanged.

Here is the same 402 images priced every way I can price it:

Read that table carefully. It does not say what an affiliate is supposed to want it to say, and it does not say the lazy opposite either.

The subscription route is not in the running for my job. $65 against $1.84 is thirty-five times the money, and the credits evaporate at month end whether I used them or not. For batch work driven by a script, a monthly credit bucket is the wrong shape of container entirely. That is not a flaw in getimg.ai; it is what subscriptions are for, and my work isn’t it.

The API route is closer than I expected. getimg.ai’s cheapest published rate, $0.015, actually undercuts Replicate’s flux-dev at $0.025 — and Z-Image Turbo is a far newer model than flux-dev. At the premium end the two platforms sit in the same band: Replicate lists flux-1.1-pro and recraft-v3 at $0.04 and ideogram-v3-quality at $0.09, against getimg.ai’s $0.03 to $0.18. The gap only blows open at the floor, where flux-schnell’s $0.003 lives and getimg.ai has nothing to answer with.

Which means the honest version is narrower than “the API wins.” It is this: for hundreds of small, cut-out, printed-at-three-inches images, the cheap old speed model is unbeatable and nothing on getimg.ai comes close. For a hundred images that someone will actually look at, the two are roughly the same money and the decision moves to which models you want and how much plumbing you feel like owning.

Confirmed, marketing, and unverified

Confirmed from getimg.ai’s own published pages today: the four plan tiers and prices; the credit allotments; that credits do not roll over; that commercial rights come only with paid plans; that API credits are separate from subscriptions and require no subscription; the published per-model API rates and the $0.015–$0.18 range; the current model roster. Those are the load-bearing facts.

Treat as marketing: any “X images per month” headline. Credit consumption varies by model, resolution, and whether you are upscaling or generating video, so a plan’s advertised image count describes a best case at the cheapest setting. Also the “all-in-one” framing — the music and speech tools are real, but a studio that does eight things is not automatically better at the one thing you need than a tool that does one.

Genuinely unverified: output quality, because I have not generated a single image there. Also unverified — and this one matters — how many credits a specific model costs in the studio. The pricing page gives totals and resolution tiers, not a per-model credit table, so you cannot budget a real project from the published numbers. And I could not confirm the current status of the free tier either way with confidence. Multiple user reports describe failed or blank generations that still consume credits; those are complaints, not measurements, and I am passing them along as complaints.

The licence line that decides whether you can sell the cover

If you are here because you make book covers or interiors, this is the paragraph that matters more than any price.

getimg.ai’s FAQ answers the commercial-use question with a qualified yes: “Yes, but only with our paid plans. Images generated under these plans can be used commercially, adhering to the CreativeML Open RAIL-M license.” Its own blog goes further, saying every paid plan includes full commercial licensing with no separate enterprise gate, and no attribution requirement for paid subscribers. So: paid plan, commercial use, RAIL-M conditions attached.

Two cautions from someone who has actually published the output of a generator. First, the RAIL-M reference is a use-restriction licence, not a blanket grant — it forbids categories of use rather than charging you for them, and it travels with the output. Read the restriction list once before you build a catalogue on it. Second, and this is the trap I wrote about in my Replicate log: licences depend on where you ran the model, not which model it was. FLUX images generated on a host can be used commercially; the same pixels generated from downloaded weights on your own machine cannot. Two paths, identical output, opposite legal answers. Whatever platform you pick, screenshot the licence terms the day you generate, and keep them with the project.

I do the same for the type and layout side, where the rules are simpler but the failure modes are just as expensive — the whole pipeline I use is in how I publish books from HTML.

The 2.0 overhaul, and why I’d price it in

One more thing a pricing table won’t tell you. getimg.ai went through a major platform overhaul this year, and it cost the company trust with its most technical users. Independent coverage reports that DreamBooth and LoRA custom-model training were discontinued on 1 March 2026, replaced by reference-image consistency tools; that the model count dropped from 80-plus community checkpoints to roughly fifteen curated models; and that users who had paid to train custom models lost them without a migration path. Multiple reviews also report that complaints were removed from the company’s feedback blog.

I have not verified those reports against the company’s own announcements, so read them as reporting rather than established fact. But the direction is consistent across several independent write-ups, and the direction is what I’d price in: this is a platform that has already removed capability customers had paid for. If your workflow depends on a specific model staying available, that is a risk you take on here and do not take when you call a model by name through an API and can swap hosts in an afternoon.

The verdict

Pick the hosted studio if you generate by hand, in a browser, in bursts — a cover concept here, a hero image there, a few dozen a month. If you want editing, inpainting, upscaling, background removal and the current flagship models behind one login without writing code, $10 or $30 a month is a fair price for not being your own infrastructure. You can see the plans here.

Stay on a raw API if you generate in bulk from a script, if your volume is spiky rather than monthly, or if your images are components — printed small, cropped, cut out, composited. My own case is the clearest possible version of this: 402 images, $1.84, a manifest file, a resumable run, and a contact sheet at the end. The subscription is thirty-five times that and the credits expire; a monthly bucket cannot beat per-call billing for work that arrives in two-day bursts.

The interesting middle: getimg.ai’s own API. If you want per-call billing and access to Seedream, Nano Banana, GPT Image and Grok Imagine behind one request format with a usage object on every response, that is a real offer at real prices, and it is not the thing the affiliate pages push. It just isn’t cheaper than the specific thing I was doing, because nothing is.

For my own next book, I’m staying on the raw API. Not out of loyalty — because $1.84 is the number I have to beat, and the way I got there was choosing the model from the output medium rather than from the marketing. That logic doesn’t change when the interface gets prettier.

The one-line version

getimg.ai is a current, well-priced hosted studio with a genuinely transparent API and a clean commercial licence on paid plans — and it is the wrong shape for scripted bulk generation, where per-call billing at $0.003 an image still wins by an order of magnitude. Buy the studio for hand-made work. Keep the API for volume. And whichever you pick, read the licence before you sell the cover.

Sources

All prices and terms checked on September 18, 2026. My own figures come from that project’s manifest file:

• getimg.ai — product site (my affiliate link; tool list and current model roster)
• getimg.ai — pricing (Entry $10 / Core $30 / Plus $65 / Ultra $175; credit allotments; 1K/2K/4K image counts; “Plan credits do not roll over”; commercial rights on paid plans under CreativeML Open RAIL-M; API credits not included in subscriptions)
• getimg.ai — image generation API (per-model rate card; “Current image rates run from $0.015 to $0.18 per image”; no subscription, no minimum spend)
• getimg.ai — developers (images from $0.015, video from $0.022 per second, usage object per response)
• replicate.com/pricing (flux-schnell $3.00 per thousand output images; flux-dev $0.025 per output image; flux-1.1-pro $0.04; recraft-v3 $0.04; ideogram-v3-quality $0.09 — re-verified today)
• Somake — getimg.ai review after the 2.0 overhaul (custom LoRA training ended 1 March 2026; model count drop; no free tier)
• The Vibe Gate — 402 images for $1.84 (my own measured run: cost breakdown, 3.5% reject rate, the licence trap)

Model rosters and rates on a platform like this change monthly, and getimg.ai has already restructured once this year. Verify anything load-bearing on the company’s own pages before you plan a budget around it. Where sources disagreed — the free tier especially — I’ve flagged the disagreement rather than picked a winner.