GPT Image 2 renders a picture from a prompt, or edits one you already have when you describe the change. Output runs 1K to 4K, priced per image, with the credit cost shown before you commit.
Three output sizes, each with its own credit cost.
Start from a prompt, or supply an image and describe the change you want.
Examples made on Kavel with this model.
Generated on this site. Every output is AI-generated, so the same prompt returns a different picture each run — read these as the range rather than as a template.
GPT Image 2 is OpenAI's image model. It renders a picture from a written prompt, and it also edits an image you supply when you describe the change rather than mask it by hand.
On Kavel it runs at this model's low-quality tier. The credit estimate is computed from the tier the request actually asks for, so the number shown before you generate is the number the run costs.
Creative engine
Per image, by output size. Pick the size in the generator and the estimate follows it.
1K, 2K and 4K are separate rates chosen before the run. A 4K render costs about four times a 1K one, so size is worth deciding after the picture is right.
Supply an image and say what should change. There is no mask to draw and no layer to manage — the instruction is the interface.
At the 1K tier a single image is a few credits, which makes trying four phrasings of the same idea an ordinary thing to do rather than a budget decision.
Describing a change works well when the change is about content — a colour, an object, a background. It is the wrong tool when you need a precise region left untouched to the pixel, because nothing here draws a mask for you.
For a different house style, Nano Banana 2 and Nano Banana Pro sit alongside it, and Seedream 5.0 Lite is the flat-rate option. When the edit is adding or replacing writing in the frame, Qwen Image 3 is the one that spells it.
Three situations where it earns the slot over the other image models here.
Describing a change beats drawing a mask when the edit is about content rather than a precise region.
At a few credits per 1K image, testing several phrasings costs less than deliberating over one.
The 4K tier is available on the same model, so the last step does not mean switching tools.
What the model accepts and returns, from this site's runtime configuration.
Describe the subject, the framing and the light, and the model renders it.
Supply a source image and describe the change; no mask, no layers.
1K, 2K and 4K, picked before the run rather than upscaled after it.
The credit cost shown reflects the exact size you selected.
A generation that does not complete is refunded automatically (unless it broke the content policy).
From the provider price table and this site's runtime config. Anything absent is absent because we could not source it.
Three steps, and the last one is the only place cost is decided.
Text-to-image invents a picture; image-to-image changes one you already have.
Name the subject, the framing and the light. Vague prompts return the most photographed version of the idea.
Settle the picture at 1K, then re-run the keeper at a larger size if it needs one.
Four jobs that suit an instruction-led image model.
Render a concept before anything is photographed, then edit details by describing them.
A few credits per image makes illustrating a post cheaper than licensing a stock photo.
Change a colour, remove a distraction, adjust a background — by saying so.
Cheap 1K passes make it reasonable to test four ideas instead of defending the first.
When a client cannot say what they want until they see it, cheap passes are worth more than one carefully argued render.
One credit pool covers Nano Banana images and Seedance and Veo video. The cost shows before every run, and failed jobs are not charged. Use a subscription for ongoing work, or a one-time pack when you just need to top up.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
One-time top-ups — buy extra credits any time you run low.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
Charges appear as “KAVEL AI” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for.
Powered by
Use one balance across every supported image and video model — eleven of them today, listed below — and check the credit cost before each request.
Video models
Image models
Model credit guide
You're only charged for successful generations. The exact estimate in the generator varies by model, length, resolution, audio, and number of images.
| Type | Model | Credit cost |
|---|---|---|
| Video | Seedance 2.5 | The long-take tier, priced per second: 5s 480p ≈ 211 credits, 5s 720p ≈ 473. A full 30s take runs ≈ 1,260 at 480p and ≈ 2,835 at 720p. Supplying a reference clip lowers the per-second rate. |
| Video | Seedance 2.0 | 5s 720p image-to-video ≈ 188 credits; text-to-video ≈ 308 credits. Scales with resolution and length. |
| Video | Seedance 2 Fast | Faster and lower cost. 5s 720p text-to-video ≈ 248 credits. |
| Video | Seedance 2 Mini | The cheapest Seedance tier. 5s 720p ≈ 154 credits, 5s 480p ≈ 72 credits. |
| Video | Seedance 1.5 Pro | Audio doubles the rate. 5s 720p ≈ 27 credits silent, ≈ 53 with audio; 1080p ≈ 57 and ≈ 113. |
| Video | Veo 3.1 | Billed per video, not per second (Lite tier). About 45 credits at 720p and 53 at 1080p. |
| Video | Kling 3.0 | Audio raises the rate. 5s 720p ≈ 105 credits silent, ≈ 150 with audio; 1080p ≈ 135 and ≈ 203. |
| Video | MiniMax H3 | Fixed 2K, no resolution ladder. Priced per second — a 5s clip ≈ 158 credits. |
| Image | Nano Banana 2 | Generate or edit from text and images. About 8 credits per 1K image, 12 at 2K, 18 at 4K. |
| Image | Nano Banana Pro | Consistent run times across generations. About 12 credits per 1K or 2K image, 21 at 4K. |
| Image | Nano Banana 2 Lite | Faster, simpler variant. A flat 8 credits per image at every resolution. Start here to test. |
| Image | GPT Image 2 | The lowest-cost image model here. About 3 credits per 1K image, 6 at 2K, 12 at 4K. |
| Image | Seedream 5.0 Lite | Flat pricing — about 8 credits per image whether you generate or edit. |
What it costs, what it accepts, and where it stops.
It bills per image by size: roughly 3 credits at 1K, 6 at 2K and 12 at 4K on this site. The estimate for the size you picked appears before you generate, and a failed run is refunded automatically (unless it broke the content policy).
The provider's low tier, which is about a third of the price of the same model elsewhere on our shelf. Higher tiers exist at the provider and are not wired here — if that changes, the rate table has to change with it, or we would bill the cheap tier for the expensive one.
Yes. Switch to the image-to-image tab, supply the source and describe the change. There is no mask to draw; the instruction does that work, which is better for content changes and worse for surgically precise regions.
Usually not. Size does not change composition, only detail, so settling the picture at 1K and re-running the keeper larger costs less than deciding at full size.
They are different house styles rather than better and worse. Nano Banana 2 and Pro sit alongside it in the picker at their own rates, and the honest way to choose is one cheap run of the same prompt through each.
Commercial rights follow your Kavel plan rather than this model, so the pricing page is where to check what your tier includes.
Write a prompt or upload an image, pick a size, and the estimate appears before you generate.
The estimate updates before you generate, and a failed run is refunded automatically (unless it broke the content policy).
