AI Image Generate

AI Image Generate API

Image generation is asynchronous:

1. POST /v1/images/generations → returns an id; credits are held, not charged
2. Wait (typically 30–60 seconds)
3. GET /v1/images/jobs/{id} until status is completed or failed
4. completed → links in data[].url. failed → credits are refunded automatically

---

Base URL and authentication

https://revidapi.com/v1

Authorization: Bearer sk_...

X-API-Key: sk_... works too. Get a key at revidapi.com/dashboard.

---

Models

curl https://revidapi.com/v1/models \
  -H "Authorization: Bearer sk_..."

Filter for entries with "kind": "image". Currently sold:

Model	Notes
gpt-image-2	Cheapest
gpt-image-2-super, gpt-image-2-4k	Same model, larger size tiers
grok-imagine-image, grok-imagine-image-quality	Grok Imagine
nano-banana, nano-banana-pro	Nano Banana
seedream-4.5, seedream-5-lite, seedream-5-pro	Seedream
flux-2-pro	Flux 2 Pro
gemini-3-pro-image, gemini-3.1-flash-image	Gemini Image
imagen-4-ultra	Imagen 4 Ultra

---

1) Create an image job

• Method: POST
• URL: https://revidapi.com/v1/images/generations
• Content-Type: application/json

Name	Type	Required	Description
model	string	Yes	From GET /v1/models
prompt	string	Yes	What the image should show
aspect_ratio	string	No	1:1, 4:3, 16:9, 9:16
size	string	Model-dependent	Size tier 1k/2k/3k/4k for Nano Banana, Seedream, Flux, Kling Image — required; omitting it returns 400 listing the tiers. GPT Image / Grok Imagine take WIDTHxHEIGHT, e.g. 2048x2048
reference_images	array	No	Image-to-image references. img_url / image_urls also accepted
output_format	string	No	png, jpeg, webp
transparent_output	bool	No	Transparent background, where the model supports it

n is not supported: the upstream returns exactly one image per billed call, so the gateway does not accept it.

Two response shapes

There are two families of image models, differing in how they return:

• Immediate (Nano Banana, Seedream, Flux, Kling Image): the response already has data[].url, no polling.
• Job-based (GPT Image, Grok Imagine): the response has id + poll; call GET /v1/images/jobs/{id}.

Write the client to handle both: if data is present use it, otherwise take id and poll.

Curl

curl -X POST https://revidapi.com/v1/images/generations \
  -H "Authorization: Bearer sk_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "a red apple on a wooden table, product photo",
    "aspect_ratio": "1:1"
  }'

Image to image

import requests

r = requests.post(
    "https://revidapi.com/v1/images/generations",
    headers={"Authorization": "Bearer sk_..."},
    json={
        "model": "nano-banana-pro",
        "prompt": "replace the background with a dark studio",
        "reference_images": ["https://example.com/product.jpg"],
        "aspect_ratio": "16:9",
    },
)
print(r.json())

Response

{
  "id": "img_xCOC2TseCCwDpyR-eqM7evGT",
  "object": "image.job",
  "status": "queued",
  "model": "gpt-image-2",
  "poll": "/v1/images/jobs/img_xCOC2TseCCwDpyR-eqM7evGT",
  "credits_reserved": 16
}

---

2) Poll the job

• Method: GET
• URL: https://revidapi.com/v1/images/jobs/{id}

curl https://revidapi.com/v1/images/jobs/img_xCOC2TseCCwDpyR-eqM7evGT \
  -H "Authorization: Bearer sk_..."

Done:

{
  "id": "img_...",
  "object": "image.job",
  "status": "completed",
  "model": "gpt-image-2",
  "waited": 45.1,
  "data": [{ "url": "https://edit.revidapi.com/media/anh/....png" }],
  "usage": { "credits_charged": 16, "credits_remaining": 352264, "seconds": 33.1 }
}

On failure status is failed and credits are already refunded.

---

Error codes

Code	Meaning	What to do
400	Missing model/prompt, or bad aspect_ratio/size	Read message
401	Wrong or missing key	Check the Authorization header
402	Not enough credits	Top up at revidapi.com/dashboard/usage
404	That model is not sold	See GET /v1/models
429	Too many concurrent images for your plan	Wait for running jobs to finish