One key,
17 models
Pay-per-use API for AI images and video. Nano Banana, GPT Image, Kling, Veo and more behind one endpoint.
Made with one key
Every picture on this page was generated through this API.
One prompt, eight models
Switch the model and see what it made from the same prompt, and what it charged.
“A small red rowing boat on a still, deep blue lake at first light, seen from directly above. One person rows; long ripples trail behind the boat across the glassy water. Photographic, natural colour.”
{"model":"nano-banana-pro","resolution":"1K"}Credits for every modelSend it yourself with one call to POST /v1/generations and your key.
Every model, one price list
Credits are charged when a request is accepted and refunded in full if it fails. Video is charged per second.
| Image model | Resolutions | Credits an image |
|---|---|---|
| nano-banana-pro11 aspect ratios | 1K · 2K · 4K | 1–2 |
| nano-banana-215 aspect ratios | 0.5K · 1K · 2K · 4K | 1–2 |
| nano-banana-2-lite11 aspect ratios | 1K | 1 |
| nano-banana8 aspect ratios | 1K | 1 |
| gpt-image-214 aspect ratios | 1K · 2K · 4K | 1–2 |
| gpt-image-2-5-flare11 aspect ratios | 1K · 2K · 4K | 1–2 |
| gpt-image-2-5-sunburst11 aspect ratios | 1K · 2K · 4K | 1–2 |
| seedream-5-pro8 aspect ratios | 1K · 2K | 1 |
| Video model | Resolutions | Credits a second |
|---|---|---|
| kling-3Clips of 3–15 s | 720p · 1080p | 1–2 |
| seedance-2Clips of 5–15 s | 480p · 720p · 1080p · 4k | 1–4 |
| seedance-2-5Clips of 4–30 s | 480p · 720p · 1080p | 1–2 |
| gemini-omniClips of 4–10 s | 720p | 1 |
| gemini-omni-flash-1-1Clips of 4–10 s | 720p · 1080p · 4k | 1–4 |
| veo-3-1Clips of 8 s | 720p · 1080p · 4k | 1–4 |
| grok-imagineClips of 6–15 s | 480p · 720p · 1080p | 1–2 |
| wan-3Clips of 2–30 s | 480p · 720p · 1080p | 1–2 |
| wan-3-primeClips of 2–30 s | 480p · 720p · 1080p | 1–2 |
How it works
Pick
17 image and video models, most served by two upstream providers.
Send
One key, one endpoint. POST the model and prompt and get an id back at once.
Collect
GET the generation for its outputUrl, a file copied to our storage that doesn't expire.
When things fail
A second provider takes over
15 of 17 models are served by two upstream providers. When the first errors, the request moves to the next without a new call from you.
Failed means free
When every route fails, the generation is marked failed with the provider's reason, and its credits go straight back to your balance.
Retries charge once
Send an Idempotency-Key header and a repeated POST returns the generation it already made instead of charging for a second one.
Renders survive our deploys
Jobs wait in a queue that outlives the workers, so a release mid-render hands the job to the next worker instead of dropping it.
Credits cost
$X.XX each
Prepaid, no subscription. One credit buys a 1K image from any image model, and video starts at one credit a second. Failed generations are refunded.
Questions
Do I need a subscription?
No. You buy credits up front and each generation spends them. There is no monthly fee and no minimum per request.
What happens when a generation fails?
First it moves to the backup provider, if the model has one. If every route fails, the generation is marked failed, you get the provider's reason, and the credits go back to your balance.
Can I retry a request safely?
Yes. Send an Idempotency-Key header. Repeating a POST with the same key and body returns the generation it already made, and charges nothing again.
How do I get the result?
Poll GET /v1/generations/{id} until its status is completed, then download from outputUrl. We copy every output to our own storage, so the link doesn't expire with the provider's.
Can I send reference images?
Yes. The image models take up to 10–16 reference images depending on the model, and most video models take reference images, clips or audio as well as a prompt.
How do I check my balance?
On the dashboard, or from code with GET /v1/balance. Every request that would overspend is refused before anything is charged.
Your first render in a few minutes
Sign in with Google, create a key and send your first request.