Stability AI

總覽

屬性	詳細資訊
說明	Stability AI 會為圖像、影片、音訊與 3D 生成建立開放 AI 模型。以 Stable Diffusion 聞名。
LiteLLM 上的提供者路由	`stability/`
提供者文件連結	Stability AI API ↗
支援的操作	`/images/generations`, `/images/edits`

LiteLLM 支援透過 Stability AI REST API（不是透過 Bedrock）進行 Stability AI 圖像生成請求。

API 金鑰

# env variable
os.environ['STABILITY_API_KEY'] = "your-api-key"

請從 Stability AI Platform 取得您的 API 金鑰。

圖像生成

使用方式 - LiteLLM Python SDK

from litellm import image_generation
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Stability AI image generation call
response = image_generation(
    model="stability/sd3.5-large",
    prompt="A beautiful sunset over a calm ocean",
)
print(response)

使用方式 - LiteLLM Proxy 伺服器

1. 設定 config.yaml

model_list:
  - model_name: sd3
    litellm_params:
      model: stability/sd3.5-large
      api_key: os.environ/STABILITY_API_KEY
    model_info:
      mode: image_generation

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY

2. 啟動 proxy

litellm --config config.yaml

# RUNNING on http://0.0.0.0:4000

3. 測試它

curl --location 'http://0.0.0.0:4000/v1/images/generations' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $LITELLM_API_KEY" \
--data '{
    "model": "sd3",
    "prompt": "A beautiful sunset over a calm ocean"
}'

進階使用方式 - 搭配額外參數

from litellm import image_generation
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

response = image_generation(
    model="stability/sd3.5-large",
    prompt="A beautiful sunset over a calm ocean",
    size="1792x1024",  # Maps to aspect_ratio 16:9
    negative_prompt="blurry, low quality",  # Stability-specific
    seed=12345,  # For reproducibility
)
print(response)

支援的參數

Stability AI 支援以下 OpenAI 相容參數：

參數	型別	說明	範例
`size`	string	圖像尺寸（對應到 aspect_ratio）	`"1024x1024"`
`n`	integer	圖像數量（注意：Stability 每次請求只回傳 1 張）	`1`
`response_format`	string	回應格式（僅 Stability 的 `b64_json`）	`"b64_json"`

尺寸到長寬比對應

size 參數會自動對應到 Stability 的 aspect_ratio：

OpenAI 尺寸	Stability 長寬比
`1024x1024`	`1:1`
`1792x1024`	`16:9`
`1024x1792`	`9:16`
`512x512`	`1:1`
`256x256`	`1:1`

使用 Stability 特定參數

您可以在請求中直接傳遞 Stability AI 專屬參數：

from litellm import image_generation
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

response = image_generation(
    model="stability/sd3.5-large",
    prompt="A beautiful sunset over a calm ocean",
    # Stability-specific parameters
    negative_prompt="blurry, watermark, text",
    aspect_ratio="16:9",  # Use directly instead of size
    seed=42,
    output_format="png",  # png, jpeg, or webp
)
print(response)

支援的圖像生成模型

模型名稱	函式呼叫	說明
sd3	`image_generation(model="stability/sd3", ...)`	Stable Diffusion 3
sd3-large	`image_generation(model="stability/sd3-large", ...)`	SD3 Large
sd3-large-turbo	`image_generation(model="stability/sd3-large-turbo", ...)`	SD3 Large Turbo（較快）
sd3-medium	`image_generation(model="stability/sd3-medium", ...)`	SD3 Medium
sd3.5-large	`image_generation(model="stability/sd3.5-large", ...)`	SD 3.5 Large（建議）
sd3.5-large-turbo	`image_generation(model="stability/sd3.5-large-turbo", ...)`	SD 3.5 Large Turbo
sd3.5-medium	`image_generation(model="stability/sd3.5-medium", ...)`	SD 3.5 Medium
stable-image-ultra	`image_generation(model="stability/stable-image-ultra", ...)`	Stable Image Ultra
stable-image-core	`image_generation(model="stability/stable-image-core", ...)`	Stable Image Core

如需更多可用模型與功能的詳細資訊，請參閱：https://platform.stability.ai/docs/api-reference

回應格式

Stability AI 會以 base64 格式回傳圖像。回應與 OpenAI 相容：

{
    "created": 1234567890,
    "data": [
        {
            "b64_json": "iVBORw0KGgo..."  # Base64 encoded image
        }
    ]
}

圖像編輯

Stability AI 支援各種圖像編輯操作，包括 inpainting、upscaling、outpainting、背景移除等。

可選參數

重要： 不同的 Stability 模型有不同的參數需求：

有些模型不需要 prompt（例如：放大、背景移除）
outpaint 模型接受數值型的 left 與 right 像素數。LiteLLM 不會轉送 up 或 down，這些是 Stability 用於垂直 outpainting 的欄位，因此目前只有水平 outpainting 可透過 stability/ 使用
目前無法透過 stability/ 使用 Style Transfer：stability/style-transfer 會解析為 Style Guide 端點（/v2beta/stable-image/control/style），且輸入一律會以 image 傳送，而 Stability 的 Style Transfer 端點需要 init_image

使用方式 - LiteLLM Python SDK

Inpainting（使用遮罩編輯）

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Inpainting - edit specific areas using a mask
response = image_edit(
    model="stability/inpaint",
    image=open("original_image.png", "rb"),
    mask=open("mask_image.png", "rb"), 
    prompt="Add a beautiful sunset in the masked area",
    size="1024x1024",
)
print(response)

圖像放大

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Conservative upscaling - preserves details
response = image_edit(
    model="stability/conservative",
    image=open("low_res_image.png", "rb"),
    prompt="Upscale this image while preserving details",
)

# Creative upscaling - adds creative details
response = image_edit(
    model="stability/creative",
    image=open("low_res_image.png", "rb"),
    prompt="Upscale and enhance with creative details",
    creativity=0.3,  # 0-0.35, higher = more creative
)

# Fast upscaling - quick upscaling (no prompt needed)
response = image_edit(
    model="stability/fast",
    image=open("low_res_image.png", "rb"),
    # No prompt required for fast upscale
)
print(response)

圖像外擴

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Extend image beyond its borders
response = image_edit(
    model="stability/outpaint",
    image=open("original_image.png", "rb"),
    prompt="Extend this landscape with mountains",
    left=100,   # Pixels to extend on the left
    right=100,  # Pixels to extend on the right
)
print(response)

背景移除

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Remove background from image
response = image_edit(
    model="stability/remove-background",
    image=open("portrait.png", "rb"),
    # No prompt required for fast upscale
)
print(response)

搜尋並取代

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Search and replace objects in image
response = image_edit(
    model="stability/search-and-replace",
    image=open("scene.png", "rb"),
    prompt="A red sports car",
    search_prompt="blue sedan",  # What to replace
)

# Search and recolor
response = image_edit(
    model="stability/search-and-recolor",
    image=open("scene.png", "rb"),
    prompt="Make it golden yellow",
    select_prompt="the car",  # What to recolor
)
print(response)

圖像控制（草圖/結構）

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Control with sketch
response = image_edit(
    model="stability/sketch",
    image=open("sketch.png", "rb"),
    prompt="Turn this sketch into a realistic photo",
    control_strength=0.7,  # 0-1, higher = more control
)

# Control with structure
response = image_edit(
    model="stability/structure",
    image=open("structure_reference.png", "rb"),
    prompt="Generate image following this structure",
    control_strength=0.7,
)
print(response)

擦除物件

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Erase objects from image
response = image_edit(
    model="stability/erase",
    image=open("scene.png", "rb"),
    mask=open("object_mask.png", "rb"),  # Mask the object to erase
    # No prompt needed
)
print(response)

Style Guide

from litellm import image_edit
import os

os.environ['STABILITY_API_KEY'] = "your-api-key"

# Generate a new image in the style of a reference image
response = image_edit(
    model="stability/style",
    image=open("style_reference.png", "rb"),  # Style reference
    prompt="A lighthouse on a cliff at sunset",
)

print(response)

支援的影像編輯模型

模型名稱	函式呼叫	說明
inpaint	`image_edit(model="stability/inpaint", ...)`	使用遮罩進行 inpainting
conservative	`image_edit(model="stability/conservative", ...)`	保守式放大
creative	`image_edit(model="stability/creative", ...)`	創意式放大
fast	`image_edit(model="stability/fast", ...)`	快速放大
outpaint	`image_edit(model="stability/outpaint", ...)`	延伸影像邊界
remove-background	`image_edit(model="stability/remove-background", ...)`	移除背景
search-and-replace	`image_edit(model="stability/search-and-replace", ...)`	搜尋並取代物件
search-and-recolor	`image_edit(model="stability/search-and-recolor", ...)`	搜尋並重新著色
sketch	`image_edit(model="stability/sketch", ...)`	以草圖控制
structure	`image_edit(model="stability/structure", ...)`	以結構控制
erase	`image_edit(model="stability/erase", ...)`	擦除物件
style	`image_edit(model="stability/style", ...)`	套用風格指南

使用方式 - LiteLLM Proxy 伺服器

1. 設定 config.yaml

model_list:
  - model_name: stability-inpaint
    litellm_params:
      model: stability/inpaint
      api_key: os.environ/STABILITY_API_KEY
    model_info:
      mode: image_edit

  - model_name: stability-upscale
    litellm_params:
      model: stability/conservative
      api_key: os.environ/STABILITY_API_KEY
    model_info:
      mode: image_edit

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY

2. 啟動 proxy

litellm --config config.yaml

# RUNNING on http://0.0.0.0:4000

3. 測試它

curl -X POST "http://0.0.0.0:4000/v1/images/edits" \
  -H "Authorization: Bearer $LITELLM_API_KEY" \
  -F "model=stability-inpaint" \
  -F "image=@original_image.png" \
  -F "mask=@mask_image.png" \
  -F "prompt=Add a beautiful garden in the masked area"

AWS Bedrock（Stability）

LiteLLM 也支援透過 AWS Bedrock 使用 Stability AI 模型。如果您已經在使用 AWS 基礎架構，這會很有幫助。

使用方式 - Bedrock Stability

from litellm import image_edit
import os

# Set AWS credentials
os.environ["AWS_ACCESS_KEY_ID"] = "your-access-key"
os.environ["AWS_SECRET_ACCESS_KEY"] = "your-secret-key"
os.environ["AWS_REGION_NAME"] = "us-east-1"

# Bedrock Stability inpainting
response = image_edit(
    model="bedrock/us.stability.stable-image-inpaint-v1:0",
    image=open("original_image.png", "rb"),
    mask=open("mask_image.png", "rb"),
    prompt="Add flowers in the masked area",
)
print(response)

# Fast upscale without prompt
response = image_edit(
    model="bedrock/stability.stable-fast-upscale-v1:0",
    image=open("low_res_image.png", "rb"),
)

# Outpaint with numeric parameters
response = image_edit(
    model="bedrock/stability.stable-outpaint-v1:0",
    image=open("original_image.png", "rb"),
    left=100,   # Automatically converted to int
    right=100,
    up=50,
    down=50,
)

print(response)

支援的 Bedrock Stability 模型

所有 Stability AI 影像編輯模型都可透過帶有 bedrock/ 前綴的 Bedrock 使用：

直接 API 模型	Bedrock 模型	說明
stability/inpaint	bedrock/us.stability.stable-image-inpaint-v1:0	Inpainting
stability/conservative	bedrock/stability.stable-conservative-upscale-v1:0	保守式放大
stability/creative	bedrock/stability.stable-creative-upscale-v1:0	創意式放大
stability/fast	bedrock/stability.stable-fast-upscale-v1:0	快速放大
stability/outpaint	bedrock/stability.stable-outpaint-v1:0	Outpainting
stability/remove-background	bedrock/stability.stable-image-remove-background-v1:0	移除背景
stability/search-and-replace	bedrock/stability.stable-image-search-replace-v1:0	搜尋並取代
stability/search-and-recolor	bedrock/stability.stable-image-search-recolor-v1:0	搜尋並重新著色
stability/sketch	bedrock/stability.stable-image-control-sketch-v1:0	以草圖控制
stability/structure	bedrock/stability.stable-image-control-structure-v1:0	以結構控制
stability/erase	bedrock/stability.stable-image-erase-object-v1:0	擦除物件

注意： Bedrock 模型 ID 可能會使用 us.stability.* 或 stability.* 前綴，視區域與模型而定。

比較路由

LiteLLM 透過兩種路由支援 Stability AI 模型：

路由	提供者	使用情境	影像生成	影像編輯
`stability/`	Stability AI Direct API	直接存取，所有最新模型	✅	✅
`bedrock/stability.*`	AWS Bedrock	AWS 整合，企業功能	✅	✅

若要直接使用 API 存取，請使用 stability/。如果您已經在使用 AWS Bedrock，請使用 bedrock/stability.*。

總覽​

API 金鑰​

圖像生成​

使用方式 - LiteLLM Python SDK​

使用方式 - LiteLLM Proxy 伺服器​

1. 設定 config.yaml​

2. 啟動 proxy​

3. 測試它​

進階使用方式 - 搭配額外參數​

支援的參數​

尺寸到長寬比對應​

使用 Stability 特定參數​

支援的圖像生成模型​

回應格式​

圖像編輯​

使用方式 - LiteLLM Python SDK​

Inpainting（使用遮罩編輯）​

圖像放大​

圖像外擴​

背景移除​

搜尋並取代​

圖像控制（草圖/結構）​

擦除物件​

Style Guide​

支援的影像編輯模型​

使用方式 - LiteLLM Proxy 伺服器​

1. 設定 config.yaml​

2. 啟動 proxy​

3. 測試它​

AWS Bedrock（Stability）​

使用方式 - Bedrock Stability​

支援的 Bedrock Stability 模型​

比較路由​

總覽

API 金鑰

圖像生成

使用方式 - LiteLLM Python SDK

使用方式 - LiteLLM Proxy 伺服器

1. 設定 config.yaml

2. 啟動 proxy

3. 測試它

進階使用方式 - 搭配額外參數

支援的參數

尺寸到長寬比對應

使用 Stability 特定參數

支援的圖像生成模型

回應格式

圖像編輯

使用方式 - LiteLLM Python SDK

Inpainting（使用遮罩編輯）

圖像放大

圖像外擴

背景移除

搜尋並取代

圖像控制（草圖/結構）

擦除物件

Style Guide

支援的影像編輯模型

使用方式 - LiteLLM Proxy 伺服器

1. 設定 config.yaml

2. 啟動 proxy

3. 測試它

AWS Bedrock（Stability）

使用方式 - Bedrock Stability

支援的 Bedrock Stability 模型

比較路由