Gemini Image Series

POST

tasks

Gemini Image Series

curl --request POST \
  --url https://www.qingbo.dev/v1/tasks \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "<string>",
  "action": "<string>",
  "prompt": "<string>",
  "n": 123,
  "seed": 123,
  "aspect_ratio": "<string>",
  "resolution": "<string>",
  "image_urls": [
    "<string>"
  ],
  "callback_url": "<string>",
  "callback_events": [
    "<string>"
  ]
}
'

import requests

url = "https://www.qingbo.dev/v1/tasks"

payload = {
    "model": "<string>",
    "action": "<string>",
    "prompt": "<string>",
    "n": 123,
    "seed": 123,
    "aspect_ratio": "<string>",
    "resolution": "<string>",
    "image_urls": ["<string>"],
    "callback_url": "<string>",
    "callback_events": ["<string>"]
}
headers = {"Content-Type": "application/json"}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

const options = {
  method: 'POST',
  headers: {'Content-Type': 'application/json'},
  body: JSON.stringify({
    model: '<string>',
    action: '<string>',
    prompt: '<string>',
    n: 123,
    seed: 123,
    aspect_ratio: '<string>',
    resolution: '<string>',
    image_urls: ['<string>'],
    callback_url: '<string>',
    callback_events: ['<string>']
  })
};

fetch('https://www.qingbo.dev/v1/tasks', options)
  .then(res => res.json())
  .then(res => console.log(res))
  .catch(err => console.error(err));

package main

import (
	"fmt"
	"strings"
	"net/http"
	"io"
)

func main() {

	url := "https://www.qingbo.dev/v1/tasks"

	payload := strings.NewReader("{\n  \"model\": \"<string>\",\n  \"action\": \"<string>\",\n  \"prompt\": \"<string>\",\n  \"n\": 123,\n  \"seed\": 123,\n  \"aspect_ratio\": \"<string>\",\n  \"resolution\": \"<string>\",\n  \"image_urls\": [\n    \"<string>\"\n  ],\n  \"callback_url\": \"<string>\",\n  \"callback_events\": [\n    \"<string>\"\n  ]\n}")

	req, _ := http.NewRequest("POST", url, payload)

	req.Header.Add("Content-Type", "application/json")

	res, _ := http.DefaultClient.Do(req)

	defer res.Body.Close()
	body, _ := io.ReadAll(res.Body)

	fmt.Println(string(body))

}

HttpResponse<String> response = Unirest.post("https://www.qingbo.dev/v1/tasks")
  .header("Content-Type", "application/json")
  .body("{\n  \"model\": \"<string>\",\n  \"action\": \"<string>\",\n  \"prompt\": \"<string>\",\n  \"n\": 123,\n  \"seed\": 123,\n  \"aspect_ratio\": \"<string>\",\n  \"resolution\": \"<string>\",\n  \"image_urls\": [\n    \"<string>\"\n  ],\n  \"callback_url\": \"<string>\",\n  \"callback_events\": [\n    \"<string>\"\n  ]\n}")
  .asString();

{
  "task_id": "task-wave1775309821b950328499",
  "model": "gemini-3-pro-image-preview",
  "action": "generate",
  "status": "queued",
  "created_at": 1775309821040,
  "progress": 0
}

Google Gemini’s native multimodal image series, three generations of Nano Banana plus dual-tier access:

Nano Banana 2 (Gemini 3.1 Flash Image Preview) — lightweight fast tier, low latency, low cost, exclusive support for Google text / image search augmentation, ideal for previewing, batches, and iteration
Nano Banana Pro (Gemini 3 Pro Image Preview) — flagship tier, native 4K output, professional fidelity, multi-step planning, built-in quality analysis
Nano Banana (Gemini 2.5 Flash Image Preview) — previous-generation workhorse, low latency, high cost-performance, multi-image fusion, character consistency

Each generation offers two tiers: standard tier (economical, value-oriented) and official direct tier (higher SLA, higher stability, higher price). The two tiers share the same model and capabilities; the difference lies in the channel and billing. We recommend selecting the corresponding -official model id directly to switch channels (clearer semantics), or you can set official_fallback: true on the standard tier to switch at runtime. Per-image billing. All versions support both Chinese and English prompts.

Pricing

Model	Price (per image)
`gemini-3-pro-image-preview-official`	`$0.1139`
`gemini-3-pro-image-preview`	`$0.0425`
`gemini-3.1-flash-image-preview-official`	`$0.05695`
`gemini-3.1-flash-image-preview`	`$0.031875`
`gemini-2.5-flash-image-preview-official`	`$0.03315`
`gemini-2.5-flash-image-preview`	`$0.01328125`

Mode Quick Reference

Mode	Trigger Fields	Supported Versions
Text-to-image	`prompt` only	All
Image-to-image	`+ image_urls` + `action: "image2image"`	All
Image editing (instruction-based)	`+ image_urls` + `action: "edit"`	All
Multi-image reference fusion	`+ image_urls` (multiple) + `action: "reference"`	All
Search augmentation (grounding)	`action: "search_grounded"` or `google_search: true`	3.1 Flash series only

Examples

curl -X POST https://www.qingbo.dev/v1/tasks \
  -H "Authorization: Bearer $WAVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-pro-image-preview",
    "prompt": "极简主义工作室人像,柔和侧光,胶片质感,4K 锐利",
    "resolution": "4K",
    "aspect_ratio": "3:2"
  }'

curl -X POST https://www.qingbo.dev/v1/tasks \
  -H "Authorization: Bearer $WAVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-pro-image-preview",
    "action": "reference",
    "prompt": "把第一张图的角色融入第二张图的场景,保持光影一致",
    "image_urls": [
      "https://cdn.example.com/character.jpg",
      "https://cdn.example.com/scene.jpg"
    ],
    "resolution": "4K",
    "aspect_ratio": "16:9"
  }'

curl -X POST https://www.qingbo.dev/v1/tasks \
  -H "Authorization: Bearer $WAVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-2.5-flash-image-preview",
    "action": "edit",
    "prompt": "把背景换成海边日落,人物保持不变",
    "image_urls": ["https://cdn.example.com/portrait.jpg"],
    "aspect_ratio": "1:1"
  }'

curl -X POST https://www.qingbo.dev/v1/tasks \
  -H "Authorization: Bearer $WAVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3.1-flash-image-preview",
    "action": "search_grounded",
    "prompt": "画一张本周登顶 App Store 的热门游戏的关键画面海报",
    "google_search": true,
    "google_image_search": true,
    "resolution": "2K",
    "aspect_ratio": "16:9"
  }'

curl -X POST https://www.qingbo.dev/v1/tasks \
  -H "Authorization: Bearer $WAVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-pro-image-preview-official",
    "prompt": "复杂排版海报,包含中英双语标语和密集小字,要求清晰可读",
    "resolution": "4K",
    "aspect_ratio": "2:3"
  }'

{
  "task_id": "task-wave1775309821b950328499",
  "model": "gemini-3-pro-image-preview",
  "action": "generate",
  "status": "queued",
  "created_at": 1775309821040,
  "progress": 0
}

After submitting, poll task status with GET /v1/tasks/{task_id}. See Task System for details.

Available Models

Model ID	Series	Description
`gemini-3-pro-image-preview-official`	Nano Banana Pro · Official Direct	Flagship tier, official direct, higher stability
`gemini-3-pro-image-preview`	Nano Banana Pro · Standard	Flagship tier, native 4K, professional fidelity, multi-step planning
`gemini-3.1-flash-image-preview-official`	Nano Banana 2 · Official Direct	Fast tier with direct channel, search augmentation
`gemini-3.1-flash-image-preview`	Nano Banana 2 · Standard	Fast tier with Google text + image search augmentation
`gemini-2.5-flash-image-preview-official`	Nano Banana · Official Direct	Previous-gen direct, high cost-performance
`gemini-2.5-flash-image-preview`	Nano Banana · Standard	Previous-gen workhorse, low latency, multi-image fusion

Common Parameters

string

required

Pick one from the Available Models list

string

default:"generate"

Operation type, allowed values:

generate — text-to-image (default)
image2image — image-to-image (requires image_urls)
edit — image editing (instruction-based / inpainting)
reference — multi-image reference fusion
search_grounded — search augmentation (3.1 Flash series only)

string

required

Image description text; supports CN/EN. The Gemini series excels at natural-language instructions and handles multi-step, multi-constraint prompts well.

integer

default:"1"

Number of images to generate (multiple per call)

integer

default:"-1"

Random seed; -1 means random. A fixed value reproduces similar results.

string

default:"1:1"

Aspect ratio. Allowed values depend on the model:Standard 10 (supported by all models):

1:1 — square
2:3 / 3:2 — portrait/landscape camera
3:4 / 4:3 — portrait/landscape standard
4:5 / 5:4 — portrait/landscape social
9:16 / 16:9 — portrait/landscape widescreen
21:9 — ultra-wide

Extreme ratios (3.1 Flash series only):

1:4 / 4:1 — tall column / banner
1:8 / 8:1 — extreme column / extreme banner

string

default:"1K"

Output resolution, depends on model:

3 Pro: 1K / 2K / 4K
3.1 Flash: 0.5K / 1K / 2K / 4K
2.5 Flash: 1K

string[]

Reference image URL array. image2image / edit typically takes one image; reference takes multiple.

string

Webhook callback URL, invoked when the task reaches a terminal state. See Callback Mechanism.

string[]

Callback event filter, e.g. ["completed", "failed"]

Model-Specific Parameters

model id: gemini-3-pro-image-preview-officialSeries: Nano Banana Pro · Official DirectSupported actions: generate / image2image / edit / referenceresolution: 1K / 2K / 4Kaspect_ratio: standard 10Highlights: Identical model to the 3 Pro standard tier, with a better channel and stability — suited for SLA-sensitive production lines. Priced ~2.7× the standard tier.

boolean

default:"false"

This model is already on the official direct tier, so this is typically unnecessary. The field is kept for upstream parameter-template compatibility.

model id: gemini-3-pro-image-previewSeries: Nano Banana Pro · StandardSupported actions: generate / image2image / edit / referenceresolution: 1K / 2K / 4Kaspect_ratio: standard 10Highlights: Flagship tier with native 4K output, professional fidelity, multi-step planning and built-in quality analysis; excels at complex composition and detailed typography.

boolean

default:"false"

Switch to the official direct tier at runtime — higher stability but higher price. We recommend selecting the -official model id directly for clearer semantics.

model id: gemini-3.1-flash-image-preview-officialSeries: Nano Banana 2 · Official DirectSupported actions: generate / image2image / edit / reference / search_groundedresolution: 0.5K / 1K / 2K / 4Kaspect_ratio: standard 10 + 4 extreme (1:4 / 4:1 / 1:8 / 8:1)Highlights: Same model as the 3.1 Flash standard tier, with a better channel and SLA.

boolean

default:"false"

Enable Google text search augmentation — the model can generate based on real-time web information.

boolean

default:"false"

Enable Google image search augmentation — uses web images as visual references.

boolean

default:"false"

This model is already on the official direct tier, so this is typically unnecessary.

model id: gemini-3.1-flash-image-previewSeries: Nano Banana 2 · StandardSupported actions: generate / image2image / edit / reference / search_groundedresolution: 0.5K / 1K / 2K / 4Kaspect_ratio: standard 10 + 4 extreme (1:4 / 4:1 / 1:8 / 8:1)Highlights: Lightweight fast tier — low latency, low cost, ideal for previewing, batches, and iteration. Exclusive support for Google text and image search augmentation, generating images grounded in real-time web information; supports extreme column / banner ratios.

boolean

default:"false"

Enable Google text search augmentation — the model can generate based on real-time web information.

boolean

default:"false"

Enable Google image search augmentation — uses web images as visual references.

boolean

default:"false"

Switch to the official direct tier at runtime. We recommend selecting the -official model id directly.

model id: gemini-2.5-flash-image-preview-officialSeries: Nano Banana · Official DirectSupported actions: generate / image2image / edit / referenceresolution: 1Kaspect_ratio: standard 10Highlights: Previous-generation Nano Banana with official direct channel — stable, suited for legacy pipelines built on 2.5 Flash.

boolean

default:"false"

This model is already on the official direct tier, so this is typically unnecessary.

model id: gemini-2.5-flash-image-previewSeries: Nano Banana · StandardSupported actions: generate / image2image / edit / referenceresolution: 1Kaspect_ratio: standard 10Highlights: Previous-generation workhorse for image generation and editing — low latency, high cost-performance, with reliable multi-image fusion and character consistency. The most economical Gemini image tier currently available.

boolean

default:"false"

Switch to the official direct tier at runtime. We recommend selecting the -official model id directly.

Resource Limits

Item	Limit
Reference images	`reference` mode recommends ≤ 4; `edit` / `image2image` typically 1
Per-image size	≤ 30MB; supports JPG / PNG / WEBP
Output resolution	See each tab (3 Pro up to 4K / 3.1 Flash up to 4K / 2.5 Flash 1K only)
Output file	JPG; URLs valid for 24 hours
Search augmentation	3.1 Flash series only; enabling `google_search` / `google_image_search` may slightly increase latency

Task System Reference — task state machine / polling cadence / async push
Request & Response — common error codes / headers / rate limits
Authentication — API key application and usage

Gemini Image Series

curl --request POST \
  --url https://www.qingbo.dev/v1/tasks \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "<string>",
  "action": "<string>",
  "prompt": "<string>",
  "n": 123,
  "seed": 123,
  "aspect_ratio": "<string>",
  "resolution": "<string>",
  "image_urls": [
    "<string>"
  ],
  "callback_url": "<string>",
  "callback_events": [
    "<string>"
  ]
}
'

import requests

url = "https://www.qingbo.dev/v1/tasks"

payload = {
    "model": "<string>",
    "action": "<string>",
    "prompt": "<string>",
    "n": 123,
    "seed": 123,
    "aspect_ratio": "<string>",
    "resolution": "<string>",
    "image_urls": ["<string>"],
    "callback_url": "<string>",
    "callback_events": ["<string>"]
}
headers = {"Content-Type": "application/json"}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

const options = {
  method: 'POST',
  headers: {'Content-Type': 'application/json'},
  body: JSON.stringify({
    model: '<string>',
    action: '<string>',
    prompt: '<string>',
    n: 123,
    seed: 123,
    aspect_ratio: '<string>',
    resolution: '<string>',
    image_urls: ['<string>'],
    callback_url: '<string>',
    callback_events: ['<string>']
  })
};

fetch('https://www.qingbo.dev/v1/tasks', options)
  .then(res => res.json())
  .then(res => console.log(res))
  .catch(err => console.error(err));

package main

import (
	"fmt"
	"strings"
	"net/http"
	"io"
)

func main() {

	url := "https://www.qingbo.dev/v1/tasks"

	payload := strings.NewReader("{\n  \"model\": \"<string>\",\n  \"action\": \"<string>\",\n  \"prompt\": \"<string>\",\n  \"n\": 123,\n  \"seed\": 123,\n  \"aspect_ratio\": \"<string>\",\n  \"resolution\": \"<string>\",\n  \"image_urls\": [\n    \"<string>\"\n  ],\n  \"callback_url\": \"<string>\",\n  \"callback_events\": [\n    \"<string>\"\n  ]\n}")

	req, _ := http.NewRequest("POST", url, payload)

	req.Header.Add("Content-Type", "application/json")

	res, _ := http.DefaultClient.Do(req)

	defer res.Body.Close()
	body, _ := io.ReadAll(res.Body)

	fmt.Println(string(body))

}

HttpResponse<String> response = Unirest.post("https://www.qingbo.dev/v1/tasks")
  .header("Content-Type", "application/json")
  .body("{\n  \"model\": \"<string>\",\n  \"action\": \"<string>\",\n  \"prompt\": \"<string>\",\n  \"n\": 123,\n  \"seed\": 123,\n  \"aspect_ratio\": \"<string>\",\n  \"resolution\": \"<string>\",\n  \"image_urls\": [\n    \"<string>\"\n  ],\n  \"callback_url\": \"<string>\",\n  \"callback_events\": [\n    \"<string>\"\n  ]\n}")
  .asString();

{
  "task_id": "task-wave1775309821b950328499",
  "model": "gemini-3-pro-image-preview",
  "action": "generate",
  "status": "queued",
  "created_at": 1775309821040,
  "progress": 0
}

Text Generation

Image Generation

Video Generation

Audio

Tools & Embeddings

Task Management

Gemini Image Series

Pricing

Mode Quick Reference

Examples

Available Models

Common Parameters

Model-Specific Parameters

Resource Limits

​Pricing

​Mode Quick Reference

​Examples

​Available Models

​Common Parameters

​Model-Specific Parameters

​Resource Limits

​Related

Pricing

Mode Quick Reference

Examples

Available Models

Common Parameters

Model-Specific Parameters

Resource Limits

Related