What is ImgAPI?
ImgAPI is a paid HTTP API and web playground that resells image generation and editing from 12 models: OpenAI's GPT Image 2, 2.5 and 2.6, Google's Nano Banana 2, 2.1 and Pro, and ByteDance's Seedream 5.0 Pro and 5.5. You send a prompt and up to 14 reference images, get a task ID back, poll until the task succeeds and then download the file. An idempotency key stops a retried request from being charged twice.
Billing is in prepaid credits bought in US dollars, with no subscription. Credits are reserved when a task starts and released if it fails, and the site pays back 20% or 100% of a task's credits when a generation runs past its time threshold. The web playground gives 3 free 1K images a day without an account; the API always uses paid credits.
The operator named in the terms is Shenzhen Sanmu Shuzhou Culture Media Co., Ltd. (深圳市三目书舟文化传媒有限公司) in Shenzhen, and the terms are governed by the laws of mainland China. A Chinese-language sister site at imgapi.vip sells the same service in yuan.
Is ImgAPI worth it?
If you are building something that makes pictures, the per-image bill is the number that decides whether the idea works. Google charges $0.134 for each Nano Banana Pro image at 1K or 2K, so a catalogue of 1,200 product shots costs about $161 before you have changed a single prompt.
ImgAPI puts a dozen image models behind one key and one request format. You submit a task, keep the ID it returns, poll until it says succeeded and download the image. The docs are careful about the boring parts: an idempotency key so a timeout does not turn into a second paid image, credits that are only settled on success, and a rule that a closed browser is not proof a task failed.
The prices are the reason anyone will look. Nano Banana Pro at 1K is $0.0215 here, about a sixth of Google's list price; Nano Banana 2.1 is $0.0143 against Google's $0.0336, and at 4K $0.0215 against $0.113. Credits are bought once in dollars and never expire.
Then read who you are buying from. The terms name a Shenzhen culture and media company, governed by mainland Chinese law. The privacy policy says prompts and reference images go to imgapi.vip, its Chinese sister site, and to image API services in mainland China, and that results may sit on Alibaba Cloud storage. ImgAPI says it does not train its own models on your content, but cannot promise every upstream provider has a no-training policy.
The catch: ImgAPI never explains how it sells below the model makers' own prices. Its Chinese site offers an "official channel" at double credits for two GPT Image 2.5 models, which suggests the default route is something else. It also sells a "GPT Image 2.6 Standard" model at 3 credits, while OpenAI's API lists only Flare and Sunburst for 2.6. There are no published rate limits, no uptime commitment and no security certification, and the terms say it does not run its own moderation system. A cheap upstream can vanish overnight, and your integration with it.
Our take: it is a market stall selling designer goods at a sixth of the shop price. The goods may be real, but you would not stock a client's launch with them.
Where it earns its place
A solo developer building a mock-up generator for a side project
Run a few thousand test images without a large bill
Submit Nano Banana 2.1 tasks at 1K for $0.0143 each, so 1,000 test images cost about $14.30 against roughly $33.60 on the Gemini API. Nothing in the prompts is private, so the routing does not matter much.
An Amazon seller with 300 products and four scenes each
Produce 1,200 lifestyle shots for listings
Banana Pro at 2K is $0.0287 an image, about $34 for the batch against about $161 at Google's rate. The product photos pass through a Chinese intermediary and Alibaba storage, which is a real risk for unreleased products.
A compliance lead at a 50-person EU marketing agency
Decide whether designers may use a cheap image API on client work
The privacy policy names mainland Chinese image services and Alibaba Cloud OSS as recipients, offers no security certification and makes no promise about upstream training. That is a no for anything under a client NDA.
Before you commit
- Setup
- Register with email or Google, buy credits, create an API key limited to chosen models with an optional expiry, then POST to /api/v1/images/generations; Node.js and Python examples poll every 5 seconds.
- Integrations
- Plain REST with bearer keys: model catalogue, credit balance, submit, task status and download endpoints. Not OpenAI-compatible, so swapping a base URL in an existing app is not enough, as the docs warn.
- Data and privacy
- Prompts and reference images go to imgapi.vip and image API services in mainland China; results may be stored on Alibaba Cloud OSS for 7 days, or 30 days for 1 extra credit. ImgAPI does not train its own models on your content, but says it cannot promise every upstream provider has a no-training policy. Which AI tools train on your data?
- Cost at scale
- 10,000 Nano Banana Pro images at 2K cost about $287 here against $1,340 at Google's standard rate; volume discounts and credit pack sizes are not published outside the signed-in dashboard.
- Limits
- Rate limits are not published beyond an HTTP 429 code; prompts up to 10,000 characters; reference files up to 10 MB each and 32 MB per request; no SLA, and model availability may change without notice.
What’s good
- Nano Banana Pro from $0.0215 an image, against $0.134 from Google
- One task-based API across 12 OpenAI, Google and ByteDance models
- Idempotency keys and settle-on-success credits guard against double charges
- Prepaid credits with no subscription; purchased credits never expire
- Timeout compensation of 20% or 100% of a task's credits
What’s not
- Does not explain how it undercuts the model makers' list prices
- Requests pass through image API services in mainland China
- No published rate limits, uptime commitment or security certification
- Sells a GPT Image 2.6 "Standard" tier that OpenAI's API does not list
- Refunds for failures and timeouts come back as credits, not cash
Best for
Hobbyists and developers prototyping image features who want the cheapest possible test volume and are not sending anything confidential.
ImgAPI alternatives
ImgAPI: common questions
- Is ImgAPI worth it?
- Image generation at a fraction of the official price, sold by a Shenzhen company that does not say how it gets those prices; fine for throwaway drafts, wrong for client work. We rate it 2.8 out of 5.
- How much does ImgAPI cost?
- Prepaid USD credits, no subscription; purchased credits do not expire. Per image: GPT Image 2.6 from $0.0054 (1K) to $0.0322 (Sunburst 4K); Nano Banana 2.1 $0.0143 (1K) to $0.0215 (4K); Nano Banana Pro $0.0215 (1K) to $0.0358 (4K); Seedream 5.5 $0.032239 to $0.044776. 3 free 1K images a day in the web playground only.
- What are the downsides of ImgAPI?
- Does not explain how it undercuts the model makers' list prices; requests pass through image API services in mainland China; no published rate limits, uptime commitment or security certification; sells a GPT Image 2.6 "Standard" tier that OpenAI's API does not list; refunds for failures and timeouts come back as credits, not cash.
- Who is ImgAPI best for?
- Hobbyists and developers prototyping image features who want the cheapest possible test volume and are not sending anything confidential.
- How much does ImgAPI cost for a team or in production?
- 10,000 Nano Banana Pro images at 2K cost about $287 here against $1,340 at Google's standard rate; volume discounts and credit pack sizes are not published outside the signed-in dashboard.
- Is ImgAPI safe for company data?
- Prompts and reference images go to imgapi.vip and image API services in mainland China; results may be stored on Alibaba Cloud OSS for 7 days, or 30 days for 1 extra credit. ImgAPI does not train its own models on your content, but says it cannot promise every upstream provider has a no-training policy.
- What are the best ImgAPI alternatives?
- The closest alternatives we have reviewed are fal, Replicate and OpenRouter.
Answers checked Nov 11, 2026.