What makes FLUX.2 [flex] different from other FLUX.2 models?

FLUX.2 [flex] is specialized for typography and text rendering, making it the best choice for preserving small details and generating text-heavy content.

Does FLUX.2 [flex] support adjustable generation steps?

Yes, FLUX.2 [flex] offers 'Quality with control' by allowing adjustable usage of steps and guidance, unlike the standard settings in [pro] and [max].

How many reference images can FLUX.2 [flex] handle?

Like the other FLUX.2 models, [flex] supports up to 8 reference images via the API (10 in Playground) for multi-reference generation.

What is the pricing for FLUX.2 [flex]?

FLUX.2 [flex] is priced at $0.06 per megapixel (MP), positioning it between the [pro] and [max] models.

Is FLUX.2 [flex] suitable for photo editing?

Yes, FLUX.2 [flex] combines powerful editing capabilities with its text specialization, allowing for changes to objects, colors, and adding text overlays sequentially.

What resolution does FLUX.2 [flex] support?

FLUX.2 [flex] supports high-resolution output up to 4 megapixels (4MP) across any aspect ratio, suitable for professional creative work.

When should I choose [flex] over [pro]?

Choose FLUX.2 [flex] when your project involves significant text elements, typography, or requires fine-grained control over generation steps that [pro] does not offer.

實惠的 FLUX 2 FLEX API | text-to-image

什麼是 FLUX.2 [flex]

FLUX.2 [flex] 是 Black Forest Labs 的 FLUX.2 系列中「靈活控制」版本：一款以生產應用為導向的文字→圖像與圖像編輯模型，提供可調整的推理參數（steps、guidance scale、seed），讓開發者能在速度與保真度之間取捨，並精準調整字體排版／細節表現。Flex 支援多參考編輯（Flex 方案最多可使用 10 張參考圖片），可編輯最高約 ~4 百萬像素的圖像。

FLUX.2 [flex] 擅長的方面

精準控制品質與速度之間的平衡（使用者可自行選擇推理步數與 guidance scale）。
優異的文字／字體排版渲染能力，相較許多同類模型更出色——推薦用於海報、UI 模擬稿、資訊圖表。
多參考編輯：可在最多 10 張輸入圖片（Flex）之間結合並保留身份／品牌元素，同時維持一致的光照與材質。
高原生解析度與編輯能力：輸出最高可達 4MP；對大多數工作流程，建議的操作範圍為 ≤2MP。
支援結構化提示詞：可使用 JSON 或自然語言提示詞；「prompt upsampling」可自動豐富簡短提示詞。

FLUX.2 [flex] 與同類模型的比較

簡表（摘要）：

Nano Banana Pro（Google／內部名稱；高保真競品）：在高解析度下具備相近的原始保真度；在某些供應商版本中，Nano Banana 可支援更大的輸出尺寸；使用者比較顯示，兩者在擬真度方面相當，但在提示詞行為與編輯一致性上各有優勢。
Qwen-Image / Hunyuan Image：根據發表時的比較，FLUX.2 在文字→圖像與編輯多參考任務上的主觀勝率更高（見發表報導）。
Midjourney / 商業 SaaS：Midjourney 常透過升頻管線來提升最終像素數以參與競爭；FLUX.2 則強調原生高解析度輸出與更緊密的多參考編輯。獨立評測者認為，兩者在構圖風格與原生像素細節上各有取捨。

實務結論：FLUX.2 的賣點在於多參考編輯、穩健的字體排版能力，以及一系列可讓團隊在吞吐量／品質／成本之間選擇折衷方案的端點。獨立測試者認為，它在同一發布窗口中的頂尖模型中具有競爭力。

[flex] 專屬參數：

Parameter	Type	Default	Description
steps	integer	50	[僅限 flex] 推理步數。最大值：50，預設值：50。數值越高 = 細節越多，但速度越慢。
guidance	float	4.5	[僅限 flex] Guidance scale。控制輸出對提示詞的遵循程度。最小值：1.5，最大值：10，預設值：4.5。數值越高 = 越貼近提示詞。

如何存取 Flux.2 Flex API

步驟 1：註冊 API 金鑰

登入 cometapi.com。若您尚未成為我們的使用者，請先註冊。登入您的 CometAPI console。取得介面的存取憑證 API key。在個人中心的 API token 處點擊「Add Token」，取得 token key：sk-xxxxx 並提交。

步驟 2：向 Flux.2 Flex API 發送請求

選擇 “flux-2-flex ”端點來發送 API 請求，並設定 request body。請求方法與 request body 可從我們網站的 API 文件取得。我們的網站也提供 Apifox 測試以方便您使用。將 <YOUR_API_KEY> 替換為您帳戶中的實際 CometAPI key。base url 為 flux generate image(https://api.cometapi.com/flux/v1/flux-2-flex)

將您的問題或請求填入 content 欄位——這就是模型將會回應的內容。處理 API 回應以取得生成的答案。

步驟 3：擷取並驗證結果

生成完成後，您可以使用 /flux/v1/get_result 端點查詢已生成的圖片或監控處理狀態。處理完成後，API 會回傳任務狀態與輸出資料。

👇 立即開始建置 Flux image generation – API 文件

FLUX 2 FLEX 的定價

探索 FLUX 2 FLEX 的競爭性定價，專為滿足各種預算和使用需求而設計。我們靈活的方案確保您只需為實際使用量付費，讓您能夠隨著需求增長輕鬆擴展。了解 FLUX 2 FLEX 如何在保持成本可控的同時提升您的專案效果。

彗星價格 (USD / M Tokens)	官方價格 (USD / M Tokens)	折扣
每次請求:$0.008	每次請求:$0	-

FLUX 2 FLEX 的範例程式碼和 API

FLUX.2 Flex 是 Black Forest Labs 的 FLUX.2 影像生成系列中面向開發者的中階成員。它會公開生成參數（尤其是採樣步數與引導尺度），讓開發者能在延遲與文字／排版保真度及影像細節之間進行取捨。

Python
JavaScript
Curl

import os
import requests

# Get your CometAPI key from https://api.cometapi.com/console/token, and paste it here
COMETAPI_KEY = os.environ.get("COMETAPI_KEY") or "<YOUR_COMETAPI_KEY>"

# API endpoint for FLUX.2 [flex]
url = "https://api.cometapi.com/flux/v1/flux-2-flex"

# Request headers
headers = {
    "Authorization": COMETAPI_KEY,
    "Content-Type": "application/json",
    "Accept": "*/*",
}

# Request payload with all available parameters
# See: https://docs.bfl.ml/api-reference/models/generate-or-edit-an-image-with-flux2-[flex]
payload = {
    # Required: Text prompt for image generation
    "prompt": "A futuristic cyberpunk cityscape at sunset, neon signs with readable text 'WELCOME TO 2077', flying cars, ultra detailed, 8k",
    # Optional: Enable prompt upsampling for enhanced details (default: true)
    "prompt_upsampling": True,
    # Output dimensions (multiple of 16, min 64, max 4MP total)
    "width": 1920,
    "height": 1080,
    # FLEX-specific: Number of inference steps (1-50, default: 50)
    # Higher values produce more detailed images but take longer
    "steps": 50,
    # FLEX-specific: Guidance scale (1.5-10, default: 5)
    # Higher values improve prompt adherence at cost of reduced realism
    "guidance": 5.0,
    # Seed for reproducible results (optional)
    "seed": 42,
    # Moderation level: 0 (strict) to 5 (permissive), default: 2
    "safety_tolerance": 2,
    # Output format: "jpeg" or "png"
    "output_format": "jpeg",
}

# Submit the generation task
print("Submitting FLUX.2 [flex] image generation task...")
response = requests.post(url, headers=headers, json=payload)
result = response.json()

print(f"Status Code: {response.status_code}")
print(f"Task ID: {result.get('id')}")
print(f"Cost: {result.get('cost')} credits")
print(f"Output Megapixels: {result.get('output_mp')}")

Python Code Example

import os
import requests

# Get your CometAPI key from https://api.cometapi.com/console/token, and paste it here
COMETAPI_KEY = os.environ.get("COMETAPI_KEY") or "<YOUR_COMETAPI_KEY>"

# API endpoint for FLUX.2 [flex]
url = "https://api.cometapi.com/flux/v1/flux-2-flex"

# Request headers
headers = {
    "Authorization": COMETAPI_KEY,
    "Content-Type": "application/json",
    "Accept": "*/*",
}

# Request payload with all available parameters
# See: https://docs.bfl.ml/api-reference/models/generate-or-edit-an-image-with-flux2-[flex]
payload = {
    # Required: Text prompt for image generation
    "prompt": "A futuristic cyberpunk cityscape at sunset, neon signs with readable text 'WELCOME TO 2077', flying cars, ultra detailed, 8k",
    # Optional: Enable prompt upsampling for enhanced details (default: true)
    "prompt_upsampling": True,
    # Output dimensions (multiple of 16, min 64, max 4MP total)
    "width": 1920,
    "height": 1080,
    # FLEX-specific: Number of inference steps (1-50, default: 50)
    # Higher values produce more detailed images but take longer
    "steps": 50,
    # FLEX-specific: Guidance scale (1.5-10, default: 5)
    # Higher values improve prompt adherence at cost of reduced realism
    "guidance": 5.0,
    # Seed for reproducible results (optional)
    "seed": 42,
    # Moderation level: 0 (strict) to 5 (permissive), default: 2
    "safety_tolerance": 2,
    # Output format: "jpeg" or "png"
    "output_format": "jpeg",
}

# Submit the generation task
print("Submitting FLUX.2 [flex] image generation task...")
response = requests.post(url, headers=headers, json=payload)
result = response.json()

print(f"Status Code: {response.status_code}")
print(f"Task ID: {result.get('id')}")
print(f"Cost: {result.get('cost')} credits")
print(f"Output Megapixels: {result.get('output_mp')}")

JavaScript Code Example

// Get your CometAPI key from https://api.cometapi.com/console/token, and paste it here
const COMETAPI_KEY = process.env.COMETAPI_KEY || "<YOUR_COMETAPI_KEY>";

// API endpoint for FLUX.2 [flex]
const url = "https://api.cometapi.com/flux/v1/flux-2-flex";

// Request payload with all available parameters
// See: https://docs.bfl.ml/api-reference/models/generate-or-edit-an-image-with-flux2-[flex]
const payload = {
    // Required: Text prompt for image generation
    prompt: "A futuristic cyberpunk cityscape at sunset, neon signs with readable text 'WELCOME TO 2077', flying cars, ultra detailed, 8k",
    
    // Optional: Enable prompt upsampling for enhanced details (default: true)
    prompt_upsampling: true,
    
    // Output dimensions (multiple of 16, min 64, max 4MP total)
    width: 1920,
    height: 1080,
    
    // FLEX-specific: Number of inference steps (1-50, default: 50)
    // Higher values produce more detailed images but take longer
    steps: 50,
    
    // FLEX-specific: Guidance scale (1.5-10, default: 5)
    // Higher values improve prompt adherence at cost of reduced realism
    guidance: 5.0,
    
    // Seed for reproducible results (optional)
    seed: 42,
    
    // Moderation level: 0 (strict) to 5 (permissive), default: 2
    safety_tolerance: 2,
    
    // Output format: "jpeg" or "png"
    output_format: "jpeg"
};

// Submit the generation task
console.log("Submitting FLUX.2 [flex] image generation task...");

const response = await fetch(url, {
    method: "POST",
    headers: {
        "Authorization": COMETAPI_KEY,
        "Content-Type": "application/json",
        "Accept": "*/*"
    },
    body: JSON.stringify(payload)
});

const result = await response.json();

console.log(`Status Code: ${response.status}`);
console.log(`Task ID: ${result.id}`);
console.log(`Cost: ${result.cost} credits`);
console.log(`Output Megapixels: ${result.output_mp}`);

Curl Code Example

#!/bin/bash
# Get your CometAPI key from https://api.cometapi.com/console/token
# Export it as: export COMETAPI_KEY="your-key-here"

curl --location --request POST 'https://api.cometapi.com/flux/v1/flux-2-flex' \
--header "Authorization: $COMETAPI_KEY" \
--header 'Content-Type: application/json' \
--header 'Accept: */*' \
--data-raw '{
    "prompt": "A futuristic cyberpunk cityscape at sunset, neon signs with readable text WELCOME TO 2077, flying cars, ultra detailed, 8k",
    "prompt_upsampling": true,
    "width": 1920,
    "height": 1080,
    "steps": 50,
    "guidance": 5.0,
    "seed": 42,
    "safety_tolerance": 2,
    "output_format": "jpeg"
}'