image-editing

图像编辑

使用 each::sense 通过自然语言指令编辑现有图像——这款智能 AI Agent 会自动为你的请求选择最佳模型。

快速开始

需要 each::labs API 密钥。请访问 eachlabs.ai 获取。

使用 curl

curl -X POST https://eachsense-agent.core.eachlabs.run/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "messages": [{"role": "user", "content": [
              {"type": "text", "text": "Change the sky to a dramatic orange and purple sunset. Keep the foreground untouched."},
              {"type": "image_url", "image_url": {"url": "https://example.com/landscape-photo.jpg"}}
            ]
          }
    ],
    "stream": false
  }'

使用 Python（OpenAI SDK）

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_EACHLABS_API_KEY",
    base_url="https://eachsense-agent.core.eachlabs.run/v1"
)

response = client.chat.completions.create(
    model="eachsense/beta",
    messages=[{"role": "user", "content": "Change the sky to a dramatic orange and purple sunset. Keep the foreground untouched."}],
    # Images are included in the message content array above
)

print(response.choices[0].message.content)

使用参考图像

curl -X POST https://eachsense-agent.core.eachlabs.run/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "messages": [{"role": "user", "content": [
              {"type": "text", "text": "Apply the color grading and mood from the reference image to the target photo"},
              {"type": "image_url", "image_url": {"url": "https://example.com/target-photo.jpg"}},
              {"type": "image_url", "image_url": {"url": "https://example.com/style-reference.jpg"}}
            ]
          }
    ],
    "stream": false
  }'

图像会以 OpenAI 多模态内容格式放入消息中发送。每个请求最多可包含 4 张图像。

流式传输

设置 "stream": true 可接收实时 SSE 响应，设置 "stream": false 则可在单个响应中获取完整结果。流式传输适合在 UI 中显示进度；非流式传输则更适合脚本和自动化。

可用模型

模型	能力	最适合的用途
flux-2-edit	自然语言图像编辑	通用编辑、风格更改、对象修改
flux-fill-pro	图像修补与扩图	填充蒙版区域、扩展画布
eachlabs-bg-remover-v1	背景移除	分离主体、创建透明背景
topaz-upscale-image	提升分辨率	编辑后放大、恢复细节
kling-face-swap	替换面部	交换不同图像中的面部

each::sense 会根据你的指令自动选择合适的图像编辑模型。用通俗的英语描述你想要的更改。

编辑能力

类别	可执行操作	指令示例
对象移除	移除不需要的元素	“移除背景中站着的人”
对象添加	添加新元素	“在长椅旁边添加一只坐着的金毛寻回犬”
风格迁移	更改艺术风格	“将这张照片转换成吉卜力工作室风格的插画”
色彩调色	调整色彩和氛围	“应用温暖的复古胶片效果，带有橙色色调”
图像修补	填充或替换区域	“把旧汽车换成一辆现代 Tesla”
扩图	扩展画布	“向左扩展这张图像，展示更多海滩”
修饰	清理瑕疵	“淡化皮肤瑕疵并提亮眼睛”
季节更换	更改时间或季节	“将场景从夏季改为下雪的冬季”
光照	修改光照条件	“将光照改为黄金时刻的效果，并带有长长的阴影”

提示词技巧

明确说明要保留的内容

始终说明哪些内容应保持不变，以避免意外修改：

"Change the wall color to teal blue. Keep the furniture, people, and floor exactly as they are."

指定具体区域

使用表示空间位置的语言来指定具体区域：

"In the top-right corner, add a flock of birds flying in V-formation"
"Replace the object on the left side of the table with a vase of sunflowers"

通过对话串联编辑

使用多轮消息依次应用编辑：

{
  "messages": [
    {"role": "user", "content": "Remove the car from the driveway"},
    {"role": "assistant", "content": "..."},
    {"role": "user", "content": [
              {"type": "text", "text": "Now change the house color to light blue"},
              {"type": "image_url", "image_url": {"url": "https://example.com/house.jpg"}},
              {"type": "image_url", "image_url": {"url": "https://example.com/landscape-with-powerlines.jpg"}}
            ]
          }
    ],
    "stream": false
  }'

风格迁移

curl -X POST https://eachsense-agent.core.eachlabs.run/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "messages": [{"role": "user", "content": [
              {"type": "text", "text": "Transform this city street photo into a Van Gogh Starry Night style painting. Keep the composition and building shapes, but apply swirling brushstrokes and vibrant blues and yellows."},
              {"type": "image_url", "image_url": {"url": "https://example.com/city-street.jpg"}}
            ]
          }
    ],
    "stream": false
  }'

图像修补（对象替换）

curl -X POST https://eachsense-agent.core.eachlabs.run/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "messages": [{"role": "user", "content": [
              {"type": "text", "text": "Replace the wooden dining table with a sleek modern glass table. Keep the chairs, plates, and everything else on the table exactly the same."},
              {"type": "image_url", "image_url": {"url": "https://example.com/dining-room.jpg"}}
            ]
          }
    ],
    "stream": false
  }'

扩图（画布扩展）

curl -X POST https://eachsense-agent.core.eachlabs.run/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "messages": [{"role": "user", "content": [
              {"type": "text", "text": "Extend this portrait photo into a full 16:9 landscape by expanding the background on both sides. Continue the indoor environment naturally with consistent lighting and style."},
              {"type": "image_url", "image_url": {"url": "https://example.com/portrait-crop.jpg"}}
            ]
          }
    ],
    "stream": false
  }'

更改色彩和氛围

curl -X POST https://eachsense-agent.core.eachlabs.run/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "messages": [{"role": "user", "content": [
              {"type": "text", "text": "Convert this daytime photo to a moody nighttime scene. Add soft neon reflections on wet pavement, streetlight glow, and a slight blue color grade. Keep all buildings and structures intact."},
              {"type": "image_url", "image_url": {"url": "https://example.com/daytime-street.jpg"}}
            ]
          }
    ],
    "stream": false
  }'

批量编辑工作流

# Apply the same edit across multiple product photos
IMAGES=(
  "https://example.com/product-1.jpg"
  "https://example.com/product-2.jpg"
  "https://example.com/product-3.jpg"
)

for IMG in "${IMAGES[@]}"; do
  curl -X POST https://eachsense-agent.core.eachlabs.run/v1/chat/completions \
    -H "Content-Type: application/json" \
    -H "X-API-Key: $EACHLABS_API_KEY" \
    -d "{
      \"messages\": [{\"role\": \"user\", \"content\": \"Remove the background and replace with a clean white studio backdrop. Add a soft drop shadow beneath the product.\"}],
      \"image_urls\": [\"$IMG\"],
      \"stream\": false
    }"
  echo "---"
done

常见问题

• 描述过于模糊：“让它更好看”会导致难以预测的结果。请明确说明要更改什么。
• 忘记说明需要保留的内容：始终说明哪些内容应保持不变，以避免不想要的编辑。
• 指令相互冲突：“移除汽车并把汽车改成红色”自相矛盾。
• 期待蒙版像素级精准：用自然语言描述区域；模型会结合上下文理解区域。
• 一次编辑过多：复杂的多步骤编辑采用连续多轮对话效果更好。

相关技能

• 图像生成 — 从头生成新图像
• 背景移除 — 专用于移除背景
• 人脸替换 — 交换照片中的面部
• 图像放大 — 编辑后提升分辨率

文档

• each::sense 概览
• each::labs API