How much does an AI image or video really cost? The exact prices, down to the cent

the real problem isn't the price, it's the fog

Prices usually hide behind opaque subscriptions and capped generations, so Imaginode displays the real compute cost of every generation, to the cent, before you click.

Ask around: how much does an AI-generated image cost? Nobody knows. The prices hide behind opaque subscriptions, capped "fast generations", waiting queues designed to push you toward the next tier up. You pay without understanding what you're buying, and you discover the bill at the end of the month.

Our conviction at Imaginode: AI creation has no reason to be a blurry subscription. Every generation has a real, known compute cost. So we might as well display it, to the cent, before you click. This article gives you all the numbers, including the uncomfortable ones, like the failure rate of video generation.

Here you'll find the price of every image model, the video prices with their two variables of duration and resolution, how refunds work, the full detail of subscriptions and top-ups, and concrete methods for spending less. By the end, you'll be able to estimate a project's budget before you've even started it. That's the whole point.

one credit is 1.2 cents, and everything becomes readable

One credit is worth about $0.012 across all 47 models: an image at 3 credits costs 3.6 cents, a video at 60 credits costs 72 cents.

The system fits in one sentence: 1 credit is worth about $0.012. An image at 3 credits therefore costs 3.6 cents, a video at 60 credits costs 72 cents. No exotic currency whose value shifts depending on the pack, no "gems" engineered to scramble your mental math. One credit, 1.2 cents, everywhere.

This single unit covers all 47 models on the platform, at the same prices whatever device you're on. It also covers the side services: a message to the assistant costs 1 credit, and so does a prompt rewrite by the magic wand. You can follow every single movement in the complete credit history, right on your profile.

Why credits rather than dollars directly? Because the amounts are tiny. Nobody wants to see "$0.058" on a button. The credit keeps the numbers whole and readable, and the conversion stays trivial: divide by one hundred and you have dollars. The poster at 31 credits costs 37 cents, period.

Balance1 250

Flux Schnell1 cr

Flux 2 Pro7 cr

Kling 2.5 · 5 s · 720p26 cr

Seedance 2.0 · 5 s · 4K330 cr

Generate · 26 cr

the price shows before you click, never after

The Generate button displays the exact cost before you click, from 1 credit per Flux Schnell attempt to 330 credits for a 4K video generation.

On Imaginode, the Generate button displays the exact cost of the generation before you click. Switch models, the number changes. Push the video from 5 to 10 seconds, the number changes. Nobody discovers the bill after the fact: it's written on the front door, not slipped under the napkin.

That interface detail changes behavior. When the cost is visible, you dare to experiment, because you know precisely what you're risking: 1 credit per attempt with Flux Schnell, which is to say practically nothing. And you think twice before the 4K generation at 330 credits, which is exactly the reflex you want to have.

And to price a project before you even open an account, the public calculator does the maths for you: pick a model, a duration, a resolution, and it shows the cost in credits and in dollars, then ranks every model from cheapest to priciest at those same settings. No sign-up, no card, just the numbers.

image prices, model by model

Images run from 1 credit with Flux Schnell to 7 credits with Flux 2 Pro, so the most expensive image on the platform costs 8.4 cents.

The entry level starts at 1 credit with Flux Schnell, perfect for roughing out an idea. GPT Image Mini follows at 2 credits. At 3 credits, you get Flux 2 Klein and Imagen 4 Fast, already very decent for social media visuals. A full iteration round of ten attempts at this level costs between 12 and 36 cents.

The mid-range plays out at 5 credits: Recraft V4.1 for design and typography, Seedream 5 Lite for recurring characters and references. The top shelf runs from 6 to 7 credits: Flux Pro 1.1 and Imagen 4 at 6, Flux Kontext at 6 for editing an existing image, Flux 2 Pro at 7 for the finals.

In other words, the most expensive image on the platform costs 8.4 cents. That number deserves a pause. It makes failure nearly free and experimentation reasonable by default. The question is no longer "can I afford one more attempt?" but "do I still have an idea worth testing?". That is the right question to be asking.

real case: a finished poster for 37 cents

Ten Flux Schnell tests at 1 credit plus three Flux 2 Pro finals at 7 credits make a print-ready poster for 31 credits, about 37 cents.

Let's walk through an actual project. You're preparing a concert poster. Exploration phase: ten attempts with Flux Schnell at 1 credit, to test compositions, moods, framings. Cost: 10 credits. You keep the best direction, adjust the prompt, then launch three final versions with Flux 2 Pro at 7 credits. Cost: 21 credits.

Total for the operation: 31 credits, roughly 37 cents for a poster worthy of being printed. Add 1 credit if the magic wand rewrote your prompt into English, 2 or 3 if the assistant helped you build the workflow. Even inflating everything, you stay under 48 cents. That is the true order of magnitude for image work.

Another frequent case: a month of posts for a professional account. Twenty published visuals, each preceded by four attempts at 1 credit and finalized with Imagen 4 at 6 credits, comes to 200 credits, so $2.40 for the month. Even doubling the iterations on the bad days, a freelancer's monthly image budget stays under $7. The barrier is no longer financial.

video costs ten times more, and that's normal

Five seconds of 720p video cost between 26 credits on Kling 2.5 Turbo and 48 on Seedance 2.0 Fast, because video demands far more compute than images.

Now for the part that stings: video. Generating 5 seconds of video demands vastly more compute than a still image, and the prices reflect that. Kling 2.5 Turbo, the platform's best value for money, runs around 26 credits for 5 seconds in 720p. Seedance 2.0 Fast, further upmarket, comes in around 48 credits for the same format.

Translate that into dollars: between 31 and 58 cents for 5 seconds in 720p. For perspective, that's the price of a hotel coffee for a shot nobody could have produced without a film crew three years ago. But video deserves more preparation than images, precisely because every failed attempt is counted in tens of cents rather than single units.

duration and resolution: the two dials that inflate the bill

Doubling the duration doubles the price, and resolution weighs even more: the same Seedance 2.0 clip goes from about 48 credits in 720p to 330 in 4K.

First dial, duration. Going from 5 to 10 seconds pushes the price up to between 60 and 100 credits depending on the model, as much as $1.20. The relationship doesn't feel linear in your head, but it is linear in the servers: twice as many frames to compute. Seedance 2.5, the high end, goes up to 30 seconds in a single take, so budget accordingly.

Second dial, resolution. The very same Seedance 2.0 generation goes from about 48 credits in 720p to about 330 credits in 4K for 5 seconds, a little over $4. So 4K is reserved for the final approved shot, never for exploration. Note also that Veo 3.1 costs more when you enable generated audio: sound is one more computation.

The method that follows from this is simple. Explore in 720p and in 5 seconds, with the cheapest model that holds up. Only raise the duration or the resolution once the movement, the framing, and the light are validated. That single reflex easily divides the video budget of a complete project by three.

when it crashes, you're refunded without asking

A technical failure on the provider's side triggers an automatic refund: the credits return to your balance within seconds, with every movement listed in your history.

Model providers have outages, saturated queues, generations that fail technically. On Imaginode, a technical failure on the provider's side triggers an automatic refund: the credits return to your balance within seconds, immediately visible. No form to fill in, no support ticket, no email to write.

Every refund appears in the credit movement history on your profile, line by line, right next to every expense. You can audit your month's consumption in a minute. That traceability seems trivial right up until the day you wonder where your 900 credits went; on that day, it's worth its weight in gold.

The typical scene: you launch a Kling video at 26 credits on a busy evening, the provider saturates, the generation fails after a minute. By the time you've reread your prompt, the 26 credits are already back in the balance shown at the top of the screen. You relaunch, and that's it. A technical failure becomes a non-event rather than a complaint.

the honest nuance: ugly but delivered still gets paid

A delivered result is paid even when it disappoints, and video disappoints roughly one time in three or four, so budget 35 to 50 credits per usable 5-second shot.

Let's be precise about what the refund does not cover: the disappointing result. If the model delivers an image, even one you consider a failure, the generation is paid for. The machine really did compute; that's the contract. Refunds based on feelings would be unworkable and, let's say it plainly, massively abused. No serious platform offers them.

This point weighs most heavily on video, where you have to be clear-eyed: roughly one generation in three or four disappoints, with the genre's classic defects, strange hands, illegible text, questionable physics. Build that rate into your budgets. A 5-second shot at 26 credits will in practice cost you more like 35 to 50 credits, failed attempts included.

This is also why the low image prices matter so much: they let you lock down everything that can be locked down before paying video rates. Composition, style, and character get validated at 1 to 5 credits per attempt. The video model then receives a start image that's already right, and its waste rate drops accordingly.

the subscriptions: 16, 50, or 170 dollars per month

Starter costs $17 for 900 credits, Creator $55 for 3,100, Studio $188 for 10,500, all before tax and cancellable anytime from the dashboard.

Three monthly plans, before tax, with credits arriving every month. Starter at $17 for 900 credits: enough to produce about thirty finished posters or a handful of short videos, ideal for discovering the platform without a heavy commitment. Creator at $55 for 3,100 credits, the heart of the range for anyone publishing every week.

And finally Studio, at $188 for 10,500 credits, designed for teams producing video continuously or serving several clients at once. Run the internal exchange rate: the Studio plan brings the credit down to $0.0162 before any promo, versus $0.0178 on Starter. The gap exists, but it stays modest: choose based on your actual volume first.

Cancellation happens at any time, from the dashboard, with no phone call and no retention form. We would rather have a customer who leaves cleanly and comes back than a trapped customer who stews. In a market where cancelling is sometimes an obstacle course, it's worth saying in black and white.

top-ups, for people who hate commitments

One-off top-ups run from 1,000 credits for $20 to 10,000 for $175, with no commitment, the credit dropping from $0.018 to $0.0158 with volume.

Don't want a subscription? One-off top-ups exist, with no commitment of any kind: 1,000 credits for $20, 2,000 for $38, 3,000 for $55, 5,000 for $90, and 10,000 for $175. The credit thus drops from $0.018 to $0.0158 depending on volume. You buy when you need it, and that's the whole story.

Top-ups suit spiky usage: a big project in March, nothing in April, two campaigns in May. The subscription, on the other hand, rewards regularity with a slightly better rate at comparable volume. Many users combine the two, a small monthly plan as working capital plus a top-up during the peaks.

One example to settle it. You estimate 2,000 credits of consumption this month, then almost nothing afterward: the $38 top-up wins, because a Creator subscription billed at $55 would leave you with 1,100 credits sitting idle. Conversely, 2,500 credits consumed every single month argues for Creator: two 1,000 top-ups plus one 2,000 top-up would cost $78, versus $55 on subscription.

the art of spending less: iterate at the floor price

Explore with Flux Schnell at 1 credit and keep premium models for the two or three finals: that sorting alone divides a project's cost by two.

The golden rule: never explore with a premium model. The first ten attempts at a visual exist to find the composition and the mood; Flux Schnell at 1 credit is more than enough for that. Save Flux 2 Pro at 7 credits or Imagen 4 at 6 for the two or three finals. That simple sorting divides a project's cost by two, with zero loss of quality at the finish line.

Second lever, the magic wand at 1 credit. A prompt rewritten into precise English by Kimi noticeably raises the rate of usable results. Spending 1 credit to avoid two failed generations at 6, the arithmetic settles itself. Same logic for the assistant: a 1-credit question that saves you from a badly built workflow is the best investment on the platform.

Third lever, the history of the last 12 generations kept on every node. Before relaunching, look at what you already have. It happens that an attempt discarded yesterday is exactly what the client is asking for today. A generation you don't relaunch is still the best price on the market: zero credits.

validate as an image before paying video rates

Correcting a mistake costs 1 to 7 credits as an image and 26 to 100 as a video, so validate framing, light and character before animating anything.

The most profitable lever of all deserves its own section. Before any video, generate the start image and validate it completely: framing, light, character, setting. An image gets corrected for 1 to 7 credits; a video that fails for the same mistake costs 26 to 100. The ratio is one to fifteen.

A costed example. You want a 10-second shot of a character walking in the rain. The spendthrift version: three 10-second Seedance 2.0 Fast videos launched on instinct, two of them failures, about 290 credits. The methodical version: five test images at 5 credits with Seedream 5 Lite, then a single video on the validated image, about 122 credits. Half the price, better result.

37 cents per poster: comparing with the world before

Against stock photos at €10 to €50 per license or shoots costing hundreds of euros, 37 cents per poster is a gap of a factor of one hundred.

Let's put these amounts in context. A single licensed photo from a classic stock library commonly bills between €10 and €50 depending on the rights. A half-day product shoot with a photographer, a studio, and retouching quickly passes several hundred euros. On the other side: 37 cents per poster, $4 for 5 seconds of 4K video.

The comparison has its limits, let's admit that. A photographer brings art direction, an eye, and a contractual responsibility that no model replaces, and some subjects genuinely demand the real thing. But for everyday visuals, posts, mockups, concept iterations, internal materials, the cost gap is not 20%: it's a factor of one hundred. These are two budgets that do not live in the same world.

which plan for which profile, and what comes next

Mostly images means Starter at $17, weekly short video means Creator at $55, teams and agencies mean Studio, and when in doubt start small and adjust.

A simple benchmark by monthly usage. You mostly produce images, a few dozen a week: Starter at $17 covers you comfortably. You publish short video every week: Creator at $55, because three or four 10-second shots per week already consume 1,200 to 1,600 credits a month. You're a small team or an agency: Studio, without hesitation.

When in doubt, start small. Take Starter, so $17 before tax, track your actual consumption on your profile for a month, then adjust: changing plans and cancelling both happen from the dashboard. The worst strategy is oversizing out of caution, paying every month for credits that just sleep. And if you'd rather decide on numbers than on a hunch, price your typical month before you pick.

The logical next step: learning to lower your waste rate, because every failed generation you avoid is worth more than any discount. Our article "writing a prompt that works" is devoted to it. And if you often create away from your desk, "creating AI images and videos on your phone" shows how to iterate at 1 credit from a train platform.