Pricing6 min readSeptember 13, 2026

GPT Image 2 Pricing: What Text-to-Image Really Costs Per Credit

Exact GPT Image 2 credit costs on VdoBloom — 3 credits at 1K, 5 at 2K, 8 at 4K — plus the two resolution rules that change your bill and what each plan buys.

GPT Image 2 text-to-image costs 3 credits at 1K, 5 credits at 2K and 8 credits at 4K on VdoBloom — about $0.15, $0.25 and $0.40 per image at the Lite subscription rate of $0.05 per credit. There is no per-prompt surcharge, no separate charge for aspect ratio, and image-to-image editing with the same model is billed at exactly the same three numbers. This page breaks down every tier, the two resolution rules that quietly change what you are billed, and what the model actually costs across our free credits, credit packs and annual plans.

The full GPT Image 2 price table

VdoBloom bills GPT Image 2 text-to-image by output resolution only, and GPT Image 2 image-to-image uses the same ladder. The credit figures below are read directly from our credit engine, and the dollar figures convert at the Lite plan rate: $15 per month buys 300 credits, so one credit is $0.05.

TierCredits per imageCost at $0.05/creditCost on annual ProWhat you get
1K text-to-image3$0.15about $0.07Social posts, thumbnails, drafts, prompt testing
2K text-to-image5$0.25about $0.12Blog heroes, ad creative, product mockups
4K text-to-image8$0.40about $0.20Print, large-format banners, crops you will zoom into
1K image-to-image3$0.15about $0.07Quick edits from a reference photo
2K image-to-image5$0.25about $0.12Retouching and restyling at usable web size
4K image-to-image8$0.40about $0.20High-resolution edits for print delivery

The annual column uses the Pro plan at its 2,000-credit tier, billed annually at $49 per month, which works out to roughly $0.0245 per credit — a little over half the monthly Lite rate, so a 1K image drops to about seven cents and a 4K image to about twenty.

Two resolution rules that change your bill

GPT Image 2 supports eleven aspect ratios on VdoBloom: auto, 1:1, 5:4, 9:16, 21:9, 16:9, 4:3, 3:2, 4:5, 3:4 and 2:3. Two of them interact with pricing, and both work in your favour rather than against it.

First, the auto aspect ratio forces a 1K output at the provider level. Our credit engine matches that behaviour: when a request arrives with auto or with no resolution set at all, it bills the 1K tier at 3 credits rather than charging you for a 2K or 4K render you were never going to receive. Second, the 1:1 square aspect ratio cannot be rendered at 4K. If you want a square image at maximum resolution, generate at 3:2 or 4:3 at 4K for 8 credits and crop, or accept 2K square for 5 credits.

The practical consequence is that a lot of GPT Image 2 usage lands on the 3-credit tier without anyone choosing it. If you want 2K or 4K, pick an explicit aspect ratio and an explicit resolution in the generator rather than leaving either on auto.

What your plan actually buys

Every new VdoBloom account starts with 10 free credits and no card. That is three 1K GPT Image 2 images with a credit left over, two 2K images, or one 4K image with two credits spare. It is enough to judge whether the model handles your subject and your text rendering before you pay anything.

Beyond that, the arithmetic is simple. A $15 Lite month at 300 credits is 100 images at 1K, 60 at 2K or 37 at 4K. The 2,000-credit Pro tier is 666 images at 1K, 400 at 2K or 250 at 4K. One-time credit packs start at $2.49 for 75 credits and never expire, which suits occasional use better than a subscription you will forget to cancel — 75 credits is 25 images at 1K. Paid plans also download watermark-free with commercial rights, which matters if these images are going into client work. Full plan details are on our pricing page.

How GPT Image 2 compares to its siblings on credits

GPT Image 2.5 arrived in September 2026 in two variants, Flare and Sunburst, and we audited both against the live provider rate before pricing them. They cost exactly the same as GPT Image 2: 3 credits at 1K, 5 at 2K, 8 at 4K, in both directions. Sunburst trades latency for precision rather than charging a premium, so there is no pricing reason to stay on GPT Image 2 if you prefer the newer models — see GPT Image 2.5 Sunburst. The one real difference is input capacity: GPT Image 2 image-to-image accepts up to 10 reference images, while 2.5 accepts 16.

Against the wider image catalogue, 3 credits is near the cheap end without being the floor. Z-image Turbo costs 2 credits a render and FLUX.2 [dev] costs 4. At the other end, Recraft V4 Pro is 22 credits and Riverflow 2.0 Pro at 4K is 27. GPT Image 2 at 8 credits for a genuine 4K output is one of the better value-per-pixel entries in the catalogue.

The filter caveat worth pricing in

GPT Image 2 carries a STRICT content label on VdoBloom, the tightest of our three tiers. That is a provider-side filter, not ours, and it is the main way people waste credits on this model: a prompt the filter dislikes is a prompt that will not produce an image. When a generation fails outright our backend refunds the deducted credits automatically, so a hard failure does not cost you. What is not refundable is a generation that succeeds but sanitises your subject into something you cannot use. For swimwear, dance, fitness and fashion work, a model on our flexible tier is the cheaper choice in practice even when its sticker price is higher — and explicit or illegal content stays blocked everywhere on the platform regardless of model.

A worked example

Say you need a 12-image product launch set: three hero shots for a landing page at 2K, six square social crops, and three 4K versions for a printed leaflet. That is 3 × 5 credits for the heroes, 6 × 3 credits for the 1K squares (square cannot go to 4K anyway), and 3 × 8 credits for the print files — 57 credits in total, or $2.85 at the Lite rate. If two of the 4K renders need a colour correction pass afterwards, that is two image-to-image edits at 8 credits each, bringing the set to 73 credits, or $3.65. Start a set in the image generator and run the corrections in the image editor.

Frequently asked questions

Is GPT Image 2 image-to-image more expensive than text-to-image?

No. Both directions cost 3 credits at 1K, 5 at 2K and 8 at 4K. Uploading reference images does not add a charge, and GPT Image 2 accepts up to 10 of them per edit.

Why was I billed 3 credits when I selected 4K?

Almost certainly because the aspect ratio was set to auto, which forces a 1K output. The billing matches the image you actually received. Pick an explicit aspect ratio and resolution to reach the higher tiers.

Do I get my credits back if a generation fails?

Yes. Our image generation and image edit endpoints refund deducted credits when a job errors out, and the refund is recorded against the original deduction so it cannot be double-applied.

Do credits bought in a pack expire?

No. One-time credit packs start at $2.49 for 75 credits and never expire. Subscription credits renew on their monthly cycle instead.

Is there a free way to try GPT Image 2?

Yes. New accounts get 10 free credits with no card, which covers three 1K images. Free-tier downloads carry a watermark; paid plans remove it and add commercial rights.

Ready to try it?

Create your first AI video in minutes — no credit card required.

Start Creating Free →