Images overview

Purpose	Endpoint
Base URL	`https://api.hop-base.com/v1`
Generate images	`POST /v1/images/generations`
Edit images	`POST /v1/images/edits`
Query async tasks	`GET /v1/images/tasks?task_id=…`

Calling GPT Image requires a key from the GPT Image (all models) group; gpt-image-2.5-flare and gpt-image-2.5-sunburst also work with a GPT Official Direct (Azure) key, returned synchronously as Base64. For other families, see "Choose a family".

Generate your first image in 3 lines

In the console's API Keys page, create a key in the GPT Image (all models) group. A key without this group (for example one with only Codex Plus or Codex Pro) returns 404 when calling gpt-image-*; you can also add the group to an existing key.

Replace sk-your-key with the new key and run. The curl example requires jq.

curl https://api.hop-base.com/v1/images/generations \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A Shiba Inu under cherry blossoms, watercolor style",
    "size": "1024x1024"
  }' \
  | jq -r '.data[0].b64_json' | base64 --decode > result.png

The image comes back as Base64 in data[].b64_json; decoded, it is a PNG. One image usually takes 15–40 seconds. Set your client's read timeout to at least 300 seconds (see GPT Image for why).

Choose a family

Families belong to different plan groups. A key that works for one does not necessarily work for another. Parameters, responses, and errors are on each family's page.

Family	Highlights	Calling mode	Reference image limit	Official price (representative model)
GPT Image	Supports mask-based local edits; billed by token	Sync; async optional	16 (official)	$5 / $30/ 1M tokens
Gemini image generation	1K–4K tiers, any aspect ratio as `WIDTHxHEIGHT`	Sync; async optional	14 (official)	$0.1344from/ image
Seedream	Returns a URL valid for 24 hours, or `b64_json`	Sync; Lite / 4.5 can stream	14	$0.045from/ image
Grok Imagine images	Does not use `size`; returns a temporary URL	Sync only	2	$0.02/ image
Kling images	Several models, including outpainting and multiple references	Async	0–10, by model	$0.0294/ image
Midjourney	Always 4 images per request	Async	1	$0.0067from/ image

Before a paid request, call GET /v1/models with the key you intend to use and take the complete model ID from the response. Each family's groups are in the quick-reference table at the top of its page.

Shared rules

Authentication and request body

Every endpoint uses Authorization: Bearer sk-your-key. The Base URL is https://api.hop-base.com/v1. The whole request body is capped at 60 MB; larger bodies return 413.

Gemini image generation also uses the OpenAI Images structure. Do not send a Gemini-native generateContent payload; HopBase adapts the request for the selected model.

Sync and async

Mode	Families	Get the result
Sync (default)	GPT Image, Gemini, Seedream, Grok	`data[]` in the response body
Async (add `Prefer: respond-async`)	GPT Image, Gemini	`GET /v1/images/tasks?task_id=…`
Always async	Kling, Midjourney	`GET /v1/video/tasks/{task_id}`

Use async for 2K / 4K images so long requests are not cut off by CDN timeouts. Seedream ignores Prefer and returns synchronously.

Pass reference images

Family	Field and shape
GPT Image	`image` on `/v1/images/edits`, or multipart `image[]`
Gemini	`image` / `images` on generations; Gemini (all models, incl. image) can also use edits
Seedream	`image` on generations; local files go through edits multipart
Grok Imagine	`image` or `image[]` on `/v1/images/edits`
Kling / Midjourney	`images: [{"url": …}]` or `[{"file_id": …}]`

URLs must be reachable by the server — not a page that requires a browser login, and not an internal network address. A Data URL must be the complete encoding with a MIME type that matches the image, for example data:image/png;base64,iVBORw0KGgo….

asset:// is only for the Seedance 2.0 asset library and cannot be used with any image family.

Billing

Actual charge = official price × group multiplier; the console's Usage records page is authoritative. Invalid-parameter 400s return before generation and are not billed. Each family's billing unit is on its page.

Prices are on each model card and in the signed-in model catalog.

Next steps

Parameters and responses by family: GPT Image, Gemini image generation, Seedream, Grok Imagine images, Kling images, Midjourney
Full parameter tables are in the API reference: Create image and Edit image
For video generation, see Video API
When something fails, check Troubleshooting

On this page