How do I handle tiered pricing?

Compute an average effective rate across the volume you expect (total charge ÷ total images or seconds) and use that as your per‑image or per‑second input. This keeps the calculator simple while still reflecting your pricing tier.

Does this include storage/bandwidth?

No. Storage, egress bandwidth, and related fees are not included. If your provider bills for them, estimate those costs separately and add them to the total.

What if I only pay per second?

Set Cost per image to zero, leave Images processed as‑is for your own reference, and enter runtime seconds and Cost per second. The calculator will then show only runtime cost contributing to the total.

Can I include image size/complexity?

Not directly. This tool assumes a single rate per image. If your provider charges more for large or complex images, estimate an average effective rate across your mix of resolutions and use that number.

Is this for training?

No. This calculator is focused on inference workloads. Training jobs typically have different pricing models and much longer runtimes, so you would need a separate calculation for training cost.

tech calculator

Image Inference Cost Calculator

Estimate inference cost for image jobs using per-image and per-second rates.

Results

Image cost: $2
Runtime cost: $0
Total cost: $2

Overview

Modern image models are often billed on two axes at once: you might pay a per‑image fee for each request to a hosted API, and also pay for GPU or container runtime by the second or minute. When you are running thousands of images per day, it is easy to lose track of the actual all‑in cost.

This image inference cost calculator helps you roll both components together. You enter the number of images, your provider’s per‑image price, any billed runtime in seconds, and the per‑second rate. The calculator then shows you the image cost, runtime cost, and combined total so you can budget batch jobs, compare providers, or sanity‑check invoices.

Use it when planning bulk generations, running evaluation suites, or budgeting recurring inference workloads for internal stakeholders.

How to use this calculator

Check your provider’s pricing page or invoice to identify the per‑image price and any time‑based billing component.
Enter the number of images you expect to process in the Images processed field.
Enter your per‑image rate in dollars in the Cost per image field. If your provider charges per request that returns multiple images, treat each request as the “image” unit or adjust accordingly.
If your provider bills GPU or container time separately, enter the total billed runtime in seconds and the per‑second rate in the Runtime (seconds) and Cost per second fields.
Review the Image cost, Runtime cost, and Total cost outputs to understand the breakdown of your expected spend.
Adjust images, rates, or runtime to explore different scenarios, such as larger batches, more efficient models, or alternative providers.

Inputs explained

Images processed: The number of images you expect to send through the model or receive as outputs during the job. For APIs that bill per image, this is typically the count used to compute per‑image charges.
Cost per image ($): Your provider’s price per image in dollars. This may be listed as a price per generated image, per input image, or per API call—choose a consistent unit and adjust the Images processed value as needed.
Runtime (seconds): Total billed compute time for GPUs, containers, or serverless functions in seconds. Some platforms bill in seconds or minutes per instance; convert to seconds and enter the total here.
Cost per second ($): The effective cost per second of runtime, in dollars. If you only know a price per minute or per hour, divide that figure by 60 or 3,600 to get a per‑second equivalent.

Outputs explained

Image cost: The total cost for all images based on the number of images processed multiplied by your per‑image rate.
Runtime cost: The total cost attributable to time‑based billing, equal to runtime seconds multiplied by your per‑second rate.
Total cost: The sum of image cost and runtime cost. This represents your estimated all‑in image inference cost for the scenario you entered, not including storage or bandwidth.

How it works

You specify how many images you plan to process or generate within a given job or time window.

You enter the per‑image rate from your provider’s pricing page—this might be a flat price per generated image, per input image, or per API call that returns one or more images.

You optionally enter the total runtime billed for your GPU, container, or serverless function in seconds, along with the per‑second rate you pay.

The calculator multiplies images by the per‑image rate to compute Image cost, and runtime seconds by the per‑second rate to compute Runtime cost.

It then adds those two components together to give you Total cost for the job at your current assumptions.

Because the math is linear, you can quickly scale scenarios up or down or plug in effective rates that already account for volume discounts or tiers.

Formula

Image cost = Images processed × Cost per image
Runtime cost = Runtime seconds × Cost per second
Total cost = Image cost + Runtime cost

If your provider quotes per‑minute or per‑hour pricing, you can convert it:
- Cost per second = (Cost per minute) ÷ 60
- Cost per second = (Cost per hour) ÷ 3,600

When to use it

Budgeting the cost of a large batch image generation job (for example, generating marketing assets, thumbnails, or dataset augmentations).
Comparing different providers or deployment options by plugging in their published per‑image and runtime prices.
Sanity‑checking your monthly inference invoices by comparing billed charges to expected jobs and runtimes.
Estimating the cost of evaluation or A/B testing runs that process many images through one or more models.
Communicating projected inference spend to product managers, finance teams, or clients in a clear, itemized way.

Tips & cautions

If your provider only charges per image and not for runtime, simply leave runtime at zero; the runtime cost will then be zero.
If you only pay for runtime (for example, self‑hosted GPUs billed by the cloud provider), set Cost per image to zero and focus on runtime inputs.
When pricing includes tiers or volume discounts, compute an average effective per‑image or per‑second rate for your expected usage level and use that here.
Remember to factor in storage, egress bandwidth, and any orchestration or platform fees separately, as those are not included in this calculator.
Re‑run this calculation when you change models, hardware, or deployment architecture, since different models and GPUs can have very different throughput and runtime profiles.
Does not explicitly model tiered, regional, or promotional pricing; it assumes a single flat rate per image and per second. Use effective average rates if your pricing is more complex.
Excludes storage, bandwidth, orchestration, and support costs; those must be added separately for a full picture of inference spend.
Assumes linear scaling of cost with images and runtime; in practice, idle time, cold starts, and concurrency limits can create overhead not captured here.
Does not differentiate between model types, resolutions, or batch sizes, which can affect throughput and thus real‑world runtime costs.
Not a billing system or invoice; actual provider charges may differ due to minimums, rounding, or additional fees.

Worked examples

Simple per‑image billing only

You plan to generate 100 images and your provider charges $0.02 per image.
Images processed = 100 and Cost per image = 0.02; you set runtime seconds to 0.
Image cost = 100 × $0.02 = $2.00.
Runtime cost = 0 × Cost per second = $0.00.
Total cost = $2.00.

Mixed billing: per‑image plus GPU runtime

You process 500 images through a self‑hosted model and also pay for GPU runtime.
Per‑image rate is $0.02, and your GPU runtime is 600 seconds at $0.0005 per second.
Image cost = 500 × $0.02 = $10.00.
Runtime cost = 600 × $0.0005 = $0.30.
Total cost ≈ $10.30 for this job.

Runtime‑only scenario on a cloud GPU

You rent a GPU at $2.16 per hour and run inference for 90 minutes, processing many images in that time.
Convert $2.16/hour to a per‑second rate: $2.16 ÷ 3,600 ≈ $0.0006 per second.
Runtime seconds = 90 minutes × 60 = 5,400 seconds; set Cost per image to $0.
Runtime cost = 5,400 × $0.0006 ≈ $3.24, which becomes your total inference compute cost.

Deep dive

Estimate image inference cost by entering images processed, per‑image rate, runtime seconds, and per‑second rate to see image cost, runtime cost, and total spend.

Use this image inference cost calculator to budget batch jobs, compare AI providers, and sanity‑check GPU or API inference bills.

FAQs

How do I handle tiered pricing?: Compute an average effective rate across the volume you expect (total charge ÷ total images or seconds) and use that as your per‑image or per‑second input. This keeps the calculator simple while still reflecting your pricing tier.
Does this include storage/bandwidth?: No. Storage, egress bandwidth, and related fees are not included. If your provider bills for them, estimate those costs separately and add them to the total.
What if I only pay per second?: Set Cost per image to zero, leave Images processed as‑is for your own reference, and enter runtime seconds and Cost per second. The calculator will then show only runtime cost contributing to the total.
Can I include image size/complexity?: Not directly. This tool assumes a single rate per image. If your provider charges more for large or complex images, estimate an average effective rate across your mix of resolutions and use that number.
Is this for training?: No. This calculator is focused on inference workloads. Training jobs typically have different pricing models and much longer runtimes, so you would need a separate calculation for training cost.

Related calculators

tech

AI Inference Cost Calculator

Estimate AI inference cost across tokens, images, or runtime seconds using per-unit rates.

tech

Token Estimator

Estimate token count from words or characters as a quick budgeting tool for LLM/API usage.

tech

PDF Token Estimator

Estimate tokens from PDF page count to budget LLM/API processing costs.

This image inference cost calculator is an estimate only and is intended for planning and educational purposes. It does not connect to your provider, account for all billing rules, or guarantee alignment with your actual invoices. Pricing, tiers, regional differences, taxes, and additional fees can materially affect real costs. Always confirm with your provider’s official pricing documentation and your own billing data before making budget, pricing, or deployment decisions.