What counts as real?
API rows use published per-call, per-image, token-derived, or per-second rates. Subscription and credit rows are labelled as estimates so they are not confused with clean API pricing.
One final image is rarely one generation. One final video is rarely one render. PixelScale translates AI creative pricing into the real project cost after retries, edits, upscales, and usable finals.
A 30 second AI video is not one click. It is seconds, retries, and renders.
Standard rates. Quality scales the unit price (representative, not exact per-provider resolution pricing).
Drag to choose what you want to keep.
$0.00
Choose a project to see the real creative workload.
Concept images, video shots, voiceover, music bed, and final assembly.
A 15-second reel can feel free, cost a few dollars, or run to thirty. Same video. The only thing that changed is how you paid for it.
Google AI Pro gives 1,000 Flow credits a month. A full-quality reel can burn hundreds once retries enter the workflow.
One clean 15-second pass on Veo 3.1, Fast to Standard, at $0.10 to $0.40 per second. No cap, you pay per second.
Creators iterate 3 to 5 times to land one usable reel. That is the figure that actually hits your card.
If you remember making video freely on a $20 plan last year, you are not misremembering. The experience felt looser, so you could iterate without thinking in per-clip costs. The newer Flow experience is more explicit about credits and compute windows. The twist: some API prices moved down while the plan still felt tighter.
Tap any provider to open its full price chart →
Trend lines sit flat at the June 2026 baseline and will fill in as prices move. Verified rates drive the comparison; subscription, credit, and audio rows are labelled estimates.
Most API image models run about $0.02 to $0.05 per generated image. Google Imagen 4 Fast is around $0.02, while FLUX 1.1 Pro and OpenAI GPT Image are around $0.04. But one usable image is rarely one generation, so a finished image usually costs more than the per-image rate.
Current API rates run roughly $0.05 to $0.10 per second of generated video. Runway Gen-4 Turbo is about $0.05 per second, Luma Dream Machine about $0.08, and Google Veo 3.1 Fast and OpenAI Sora 2 about $0.10 per second at 720p. Higher resolutions and added audio cost more.
Because you rarely keep the first result. Landing one usable image often takes several attempts, plus edits or upscales. PixelScale estimates the realistic project cost after those retries, not just the sticker rate per generation.
At list prices, Google Imagen 4 Fast (around $0.02 per image) is currently the lowest among the major API providers PixelScale tracks. Subscription tools like Midjourney can work out cheaper per image at high volume, but they are harder to compare directly.
Yes. PixelScale is a free tool from Bilton Projects, with no sign-up and no tracking. Prices are verified against official provider documentation where possible and clearly labelled as estimates otherwise.
Rates are checked against official pricing pages and dated in the changelog. Creative-AI prices change often, so PixelScale tracks notable changes and shows when each rate was last verified.
API rows use published per-call, per-image, token-derived, or per-second rates. Subscription and credit rows are labelled as estimates so they are not confused with clean API pricing.
Creative pricing changes fast. Send the provider, model, rate, source link, and what looks wrong.
Report pricing errorThe default view assumes retries. That is where creative AI costs become real: ten usable images can require eighty attempts. The calculator shows both the floor and the realistic range.