AI Image Generation Cost Calculator — Price per Image

Quick Start Guide

  1. Pick a service & model: Search the model catalog — OpenAI, Google, Midjourney, Stability AI, Black Forest Labs, Ideogram, Recraft, and xAI — grouped by provider with resolution shown per model.
  2. Choose Simple or Compare mode: Simple estimates one model at a time. Compare benchmarks the same workload across several models side by side.
  3. Enter requested images: Set your target final image count — the number of usable images you actually need, before retries or variations.
  4. Optionally load a campaign preset: Prefill Run Assumptions with a typical workload (social ads, ecommerce catalog, blog thumbnails, and more), or leave it collapsed to use the defaults.
  5. Tune run assumptions: Adjust variations per prompt, retry rate, and runs per month to match how you actually generate images.
  6. Review the cost: See cost per image, total run cost, and — if runs per month is set — the projected monthly spend.

Understanding Image Generation Pricing

Why the same model has several prices

Most providers price by resolution and quality, not by a single flat per-image rate. A larger image or a higher "quality" setting (e.g. gpt-image-1 Low/Medium/High, or Imagen Standard vs Ultra) costs more because it takes more compute to generate — effectively a megapixel-based charge, even when the provider lists it as separate named tiers rather than a per-megapixel formula. Selecting the right model tier below already accounts for this; there is no separate "resolution surcharge" to add.

Retries and variations multiply cost

A prompt rarely produces a usable image on the first try. Run Assumptions lets you model a retry rate (failed generations you regenerate) and variations per prompt (multiple options per idea) — both multiply the effective image count, and therefore the bill, well above the raw number of prompts you write.

Monthly spend vs. one-off run cost

A single run cost tells you what one batch of images costs today. Runs per month projects that same batch recurring — useful for budgeting an ongoing content pipeline (e.g. a weekly product-photo batch) rather than a one-time project.

Features

8-Provider Model Catalog: OpenAI, Google, Midjourney, Stability AI, Black Forest Labs, Ideogram, Recraft, and xAI, with resolution and pricing tier shown per model.

Compare Mode: Benchmark the same workload across up to seven additional models at once, with the lowest-cost option highlighted.

Campaign Presets: Six ready-made workload templates — social ad batch, ecommerce catalog, blog thumbnails, concept art, pitch deck visuals, and weekly content — prefill realistic variations, retry rate, and run cadence.

Run Assumptions: Model variations per prompt, retry overhead, and monthly run cadence, plus a token-aware pricing basis for models billed by generation token count.

Custom Pricing: Override the catalog rate with your own negotiated contract price per image.

Export Data: Export your image generation cost calculations to JSON, CSV, or PDF format for project planning and budgeting.

Common Use Cases

Social Ad Batch: 30 images with 3 variations per prompt and a 15% retry rate, refreshed 4 times a month — a typical paid-social creative cadence.

Ecommerce Catalog: 120 product images with 2 variations and a 10% retry rate, refreshed twice a month — sized for a recurring catalog update.

Blog Thumbnail Pack: 20 images per batch, refreshed weekly — recurring blog and SEO thumbnail production at low retry overhead.

Concept Art Sprint: 80 images with 4 variations per prompt and a 20% retry rate for one high-iteration ideation cycle — comparing models for the best cost-to-quality tradeoff.

Frequently Asked Questions

The calculator multiplies your requested image count by variations per prompt and retry overhead to get an effective image count, then applies the selected model's price per image. If runs per month is set, monthly cost is that per-run cost multiplied by run cadence.
Compare mode evaluates the same workload assumptions (images, retries, variations, run cadence) across up to eight selected models and shows which model has the lowest projected cost.
It varies by provider, resolution, and quality tier — pricing spans OpenAI, Google, Midjourney, Stability AI, Black Forest Labs, Ideogram, Recraft, and xAI, each with several tiers. Use Compare mode to run your own image count and retry assumptions across several models at once and see them ranked by projected cost.
Yes, for most use cases. A single AI-generated image typically costs a few cents to a few tenths of a dollar depending on the model and quality tier, while hiring a designer typically costs $50-200+ per image. For bulk generation (100+ images), AI is significantly more cost-effective, though custom branding and specific requirements may still need human designers.
Per Image is the straightforward one: each provider publishes a flat price per finished image, and the calculator multiplies it by how many images your run produces. Token-Aware exists because some providers meter generation by tokens rather than by the finished picture, so the headline per-image figure understates what you are actually billed. Selecting it adds a flat 12% on top as a rough allowance. Treat that as an approximation and nothing more — it is applied to every model equally rather than derived from any one provider's token rates. If you know your real cost, the custom pricing override will always beat it.
Yes, and on some models it matters more than the provider does. Within one family the tiers in this catalog range from roughly twice the price to fifteen times it — GPT Image 1 runs $0.011 to $0.167 an image across its tiers, while Stable Diffusion 3.5 spans a much narrower $0.035 to $0.065. Comparing different providers at the same tier is wide as well, so neither choice reliably dominates the other; the only way to know is to price your own run both ways. The model picker shows resolution and tier next to each entry so you can compare like for like, and Compare mode ranks whichever models you pick by projected cost.
Model prices come from each provider's published pricing pages and the date they were last checked is shown above the model picker. A handful of models that providers have not published a rate for carry a marked estimate instead — those are labelled "est." in their name, so you can see at a glance which figures are firm. Deprecated or retired models are grouped as legacy in the picker. If your contract rate differs from the catalog default, use the custom pricing override to enter your own per-image cost.

Last updated