知识库首页 seo-llm 资料 AI-IMAGE-GENERATION-GUIDE.md

AI IMAGE GENERATION GUIDE

本地来源：seo-llm/raw/seo知识库/seo方法论/seo-knowledge-base/Guideline/AI-IMAGE-GENERATION-GUIDE.md

AI 图片生成指南 - Kie.ai API

目标: 使用 Kie.ai API 为博客文章生成高质量配图 优先模型: Nano Banana (快速、高质量、成本低) 备用模型: Midjourney (超高质量、风格多样)

📋 目录

快速开始
API Key 配置
Nano Banana 模型使用
Midjourney 模型使用
实战示例：博客配图生成
最佳实践
故障排查

🚀 快速开始

为什么使用 AI 生图？

✅ 快速高效: 30秒-2分钟生成高质量图片
✅ 成本低廉: 比购买素材或雇设计师便宜
✅ 版权无忧: AI 生成图片无版权纠纷
✅ 风格统一: 保持品牌视觉一致性

模型选择

模型	速度	质量	成本	适用场景
Nano Banana ⭐	快 (30秒)	高	20积分	博客配图、社交媒体、快速原型
Nano Banana Edit	快 (30秒)	高	20积分	图片编辑、风格迁移
Midjourney Relaxed	中 (1-2分钟)	超高	30积分	精美配图、封面图、展示图
Midjourney Fast	快 (30秒)	超高	80积分	紧急需求、高质量配图
Midjourney Turbo	极快 (15秒)	超高	160积分	紧急高质量需求

推荐策略:

日常博客配图 → Nano Banana
重要文章封面 → Midjourney Relaxed
紧急高质量需求 → Midjourney Fast

🔑 API Key 配置

步骤 1: 获取 API Key

访问 Kie.ai API Key 管理页面
登录账号
点击 "生成 API Key"
复制 API Key（格式: kie_xxxxxxxxxxxxxxxxxxxxxxxx）

步骤 2: 配置环境变量

# 在项目根目录的 .env 文件中添加
KIE_API_KEY=kie_xxxxxxxxxxxxxxxxxxxxxxxx

安全提醒:

❌ 不要将 API Key 提交到 Git
✅ .env 文件已在 .gitignore 中
✅ 定期轮换 API Key

步骤 3: 验证配置

# 测试 API Key 是否有效
curl -H "Authorization: Bearer $KIE_API_KEY" \
  https://api.kie.ai/api/v1/jobs/createTask

🍌 Nano Banana 模型使用

模型特点

速度: 30秒生成
质量: 高质量，适合博客配图
成本: 20积分/图
分辨率: 支持多种比例（1:1, 16:9, 9:16 等）
格式: PNG/JPEG

基础用法

1. Text-to-Image (文字生图)

# Step 1: 创建生成任务
curl -X POST https://api.kie.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/nano-banana",
    "input": {
      "prompt": "A futuristic AI laboratory with glowing neural networks, holographic displays showing video upscaling process, sleek modern interior, vibrant blue and purple lighting, photorealistic, 4K",
      "image_size": "16:9",
      "output_format": "png"
    }
  }'

# 响应示例
{
  "code": 200,
  "msg": "success",
  "data": {
    "taskId": "281e5b0*********************f39b9"
  }
}

# Step 2: 查询任务结果
curl -X GET "https://api.kie.ai/api/v1/jobs/recordInfo?taskId=281e5b0*********************f39b9" \
  -H "Authorization: Bearer YOUR_API_KEY"

# 成功响应
{
  "code": 200,
  "msg": "success",
  "data": {
    "taskId": "281e5b0*********************f39b9",
    "state": "success",
    "resultJson": "{\"resultUrls\":[\"https://file.aiquickdraw.com/...\"]}"
  }
}

2. Image-to-Image (图片编辑)

# 使用 Nano Banana Edit 模型
curl -X POST https://api.kie.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/nano-banana-edit",
    "input": {
      "prompt": "Transform this into a vibrant digital art style with neon colors and cyberpunk aesthetic",
      "image_urls": ["https://example.com/original-image.jpg"],
      "image_size": "16:9",
      "output_format": "png"
    }
  }'

图片比例选择

比例	用途	尺寸参考
`1:1`	社交媒体、缩略图	1024x1024
`16:9`	博客封面、横幅图	1920x1080
`9:16`	手机竖屏、Stories	1080x1920
`4:3`	传统显示器	1600x1200
`3:2`	照片打印	1800x1200

博客配图推荐: 16:9 (最常用)

Node.js 代码示例

// scripts/generate-nano-banana-image.js
const axios = require('axios');

const KIE_API_KEY = process.env.KIE_API_KEY;
const API_BASE_URL = 'https://api.kie.ai/api/v1/jobs';

async function generateImage(prompt, imageSize = '16:9') {
  try {
    // Step 1: 创建任务
    const createResponse = await axios.post(
      `${API_BASE_URL}/createTask`,
      {
        model: 'google/nano-banana',
        input: {
          prompt,
          image_size: imageSize,
          output_format: 'png',
        },
      },
      {
        headers: {
          Authorization: `Bearer ${KIE_API_KEY}`,
          'Content-Type': 'application/json',
        },
      }
    );

    const taskId = createResponse.data.data.taskId;
    console.log(`✅ 任务创建成功，Task ID: ${taskId}`);

    // Step 2: 轮询查询结果
    let attempts = 0;
    const maxAttempts = 60; // 最多等待 60 * 5 = 300秒 (5分钟)

    while (attempts < maxAttempts) {
      await new Promise((resolve) => setTimeout(resolve, 5000)); // 等待 5 秒

      const resultResponse = await axios.get(
        `${API_BASE_URL}/recordInfo?taskId=${taskId}`,
        {
          headers: {
            Authorization: `Bearer ${KIE_API_KEY}`,
          },
        }
      );

      const { state, resultJson } = resultResponse.data.data;

      if (state === 'success') {
        const result = JSON.parse(resultJson);
        console.log(`🎉 图片生成成功！`);
        console.log(`📷 图片地址: ${result.resultUrls[0]}`);
        return result.resultUrls[0];
      } else if (state === 'fail') {
        throw new Error('图片生成失败');
      }

      attempts++;
      console.log(`⏳ 等待中... (${attempts}/${maxAttempts})`);
    }

    throw new Error('任务超时');
  } catch (error) {
    console.error('❌ 错误:', error.message);
    throw error;
  }
}

// 使用示例
const prompt =
  'A modern AI-powered video upscaling interface, sleek dark theme, glowing blue accents, before/after comparison sliders, futuristic tech aesthetic, professional software UI, 4K quality';
generateImage(prompt, '16:9').then((imageUrl) => {
  console.log('✅ 生成完成，图片 URL:', imageUrl);
});

运行:

node scripts/generate-nano-banana-image.js

🎨 Midjourney 模型使用

模型特点

速度: Relaxed (1-2分钟), Fast (30秒), Turbo (15秒)
质量: 超高质量，艺术感强
成本: Relaxed (30积分), Fast (80积分), Turbo (160积分)
特色: 支持风格参考、图片编辑、视频生成

基础用法

1. Text-to-Image

curl -X POST https://api.kie.ai/api/v1/mj/generate \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskType": "mj_txt2img",
    "prompt": "A cinematic shot of advanced AI technology upscaling vintage home videos, holographic displays showing before/after comparison, warm nostalgic lighting mixed with futuristic elements, photorealistic, 8K --ar 16:9 --v 7",
    "speed": "relaxed",
    "aspectRatio": "16:9",
    "version": "7"
  }'

2. Image-to-Image

curl -X POST https://api.kie.ai/api/v1/mj/generate \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskType": "mj_img2img",
    "prompt": "Transform into a vibrant digital illustration with sci-fi elements",
    "fileUrls": ["https://example.com/source-image.jpg"],
    "speed": "relaxed",
    "aspectRatio": "16:9",
    "version": "7"
  }'

3. 查询任务结果

curl -X GET "https://api.kie.ai/api/v1/mj/record-info?taskId=mj_task_abcdef123456" \
  -H "Authorization: Bearer YOUR_API_KEY"

# 成功响应（通常生成 4 张图片）
{
  "code": 200,
  "data": {
    "successFlag": 1,
    "resultInfoJson": {
      "resultUrls": [
        { "resultUrl": "https://tempfile.aiquickdraw.com/v/image_0.jpeg" },
        { "resultUrl": "https://tempfile.aiquickdraw.com/v/image_1.jpeg" },
        { "resultUrl": "https://tempfile.aiquickdraw.com/v/image_2.jpeg" },
        { "resultUrl": "https://tempfile.aiquickdraw.com/v/image_3.jpeg" }
      ]
    }
  }
}

4. 放大图片 (Upscale)

# 选择 4 张图片中的一张进行放大（imageIndex: 0-3）
curl -X POST https://api.kie.ai/api/v1/mj/generateUpscale \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskId": "mj_task_abcdef123456",
    "imageIndex": 0
  }'

Midjourney Prompt 技巧

基础结构

[主体] + [风格] + [细节] + [参数]

示例 Prompts

博客封面:

A professional tech blog header featuring AI-powered video restoration, split-screen showing old grainy footage transforming into crystal-clear 4K, modern minimalist design, cool blue gradient background, glass morphism UI elements --ar 16:9 --v 7 --stylize 100

教程配图:

Step-by-step tutorial illustration of AI upscaling workflow, clean infographic style, numbered steps with icons, arrows showing process flow, vibrant gradient accents, white background, professional and easy to understand --ar 16:9 --v 7

产品展示:

Sleek product showcase of AI video upscaling software interface on MacBook Pro, dark mode UI with neon blue accents, before/after video comparison playing, modern office desk setup, soft studio lighting, photorealistic --ar 16:9 --v 7 --stylize 200

Midjourney 参数说明

参数	说明	范围	推荐值
`--ar`	宽高比	1:1, 16:9, 9:16 等	16:9 (博客)
`--v`	模型版本	5.1, 6, 7	7 (最新)
`--stylize`	风格化程度	0-1000	100 (自然), 500 (艺术)
`--chaos`	多样性	0-100	10 (稳定), 50 (多样)
`--weird`	怪异度	0-3000	0 (正常), 500 (创意)

💡 实战示例：博客配图生成

场景 1: 为 "AI 视频放大教程" 生成封面图

需求:

16:9 横幅图
展示 AI 技术感
体现视频质量提升

Nano Banana 方案 (快速、性价比高):

// scripts/generate-blog-cover.js
const prompt = `
A modern AI video upscaling interface, split-screen comparison showing blurry old video on left transforming into sharp 4K quality on right,
sleek dark software UI with glowing blue progress bar,
futuristic holographic elements, professional tech aesthetic,
cinematic lighting, photorealistic, 8K quality
`;

generateImage(prompt, '16:9');

Midjourney 方案 (高质量):

curl -X POST https://api.kie.ai/api/v1/mj/generate \
  -H "Authorization: Bearer $KIE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskType": "mj_txt2img",
    "prompt": "Cinematic AI video upscaling demonstration, vintage home movie transforming into crystal-clear 4K, holographic interface overlays, warm nostalgic colors mixed with futuristic blue tech elements, emotional and technological, photorealistic --ar 16:9 --v 7 --stylize 200",
    "speed": "relaxed",
    "aspectRatio": "16:9",
    "version": "7",
    "stylization": 200
  }'

场景 2: 为 "ComfyUI 工作流指南" 生成配图

需求:

清晰的流程图
专业技术风格
易于理解

Prompt:

Clean technical diagram of ComfyUI workflow for video upscaling,
node-based visual programming interface,
connected blocks showing input → processing → output flow,
minimalist dark theme with neon green connection lines,
professional software documentation style,
highly readable, infographic quality --ar 16:9 --v 7

场景 3: 批量生成系列文章配图

// scripts/batch-generate-images.js
const blogTopics = [
  {
    title: 'Ultimate Upscaling Guide 2026',
    prompt:
      'Professional AI upscaling software interface showcasing various enhancement options, before/after quality comparison sliders, modern dark UI design --ar 16:9',
  },
  {
    title: 'Low VRAM Workflow',
    prompt:
      'Computer hardware diagram highlighting GPU memory optimization, visual representation of efficient resource usage, technical illustration style --ar 16:9',
  },
  {
    title: '8GB VRAM Secrets',
    prompt:
      'Infographic showing memory management tips for AI video processing, clean layout with icons and statistics, professional tech aesthetic --ar 16:9',
  },
];

async function batchGenerate() {
  for (const topic of blogTopics) {
    console.log(`\n🎨 生成配图: ${topic.title}`);
    const imageUrl = await generateImage(topic.prompt);
    console.log(`✅ 完成: ${imageUrl}\n`);

    // 等待 5 秒避免 API 限流
    await new Promise((resolve) => setTimeout(resolve, 5000));
  }
}

batchGenerate();

📐 最佳实践

1. Prompt 编写技巧

✅ 好的 Prompt

A modern AI-powered video editing interface displaying real-time upscaling progress,
sleek dark theme with blue accent colors,
split-screen comparison showing SD to 4K transformation,
professional software UI design,
clean and minimalist,
photorealistic rendering,
16:9 aspect ratio

特点:

具体描述主体
明确风格和色调
包含技术细节
指定质量要求

❌ 差的 Prompt

video upscaling software

问题:

过于简单
缺少细节
没有风格指导
结果不可预测

2. 图片比例选择

用途	推荐比例	说明
博客封面	16:9	标准横幅，适配大多数网站
社交媒体	1:1	Instagram, Twitter 缩略图
Stories	9:16	Instagram/Facebook Stories
打印海报	3:2 或 4:3	传统打印比例

3. 成本优化

策略 1: 优先使用 Nano Banana

日常配图用 Nano Banana (20积分)
重要封面用 Midjourney Relaxed (30积分)
紧急情况用 Midjourney Fast (80积分)

策略 2: 批量生成

// 一次生成多张，选择最佳
const response = await axios.post(API_URL, {
  model: 'google/nano-banana',
  input: {
    prompt: yourPrompt,
    num_images: 4, // Nano Banana 支持多图
  },
});

策略 3: 复用和编辑

保存满意的图片作为参考
使用 Image-to-Image 进行微调
建立 Prompt 模板库

4. 图片文件管理

# 下载并保存图片的脚本
# scripts/download-generated-image.js

const fs = require('fs');
const path = require('path');
const https = require('https');

async function downloadImage(url, filename) {
  const filepath = path.join(__dirname, '../public/blog-images', filename);

  return new Promise((resolve, reject) => {
    https.get(url, (response) => {
      const fileStream = fs.createWriteStream(filepath);
      response.pipe(fileStream);

      fileStream.on('finish', () => {
        fileStream.close();
        console.log(`✅ 图片已保存: ${filepath}`);
        resolve(filepath);
      });

      fileStream.on('error', reject);
    });
  });
}

// 使用示例
downloadImage(
  'https://file.aiquickdraw.com/custom-page/akr/section-images/1756223371764w82dsmi4.png',
  'tutorial-ultimate-upscaling-guide-2026.png'
);

5. 图片优化

生成后的图片建议进行优化：

# 使用 sharp 库优化图片
npm install sharp

# scripts/optimize-image.js
const sharp = require('sharp');

async function optimizeImage(inputPath, outputPath) {
  await sharp(inputPath)
    .resize(1920, 1080, { fit: 'inside' })  // 限制最大尺寸
    .webp({ quality: 85 })                   // 转换为 WebP 格式
    .toFile(outputPath);

  console.log(`✅ 图片已优化: ${outputPath}`);
}

optimizeImage(
  'public/blog-images/original.png',
  'public/blog-images/optimized.webp'
);

🔧 故障排查

问题 1: API Key 无效

错误信息:

{
  "code": 401,
  "msg": "Unauthorized - Authentication credentials are missing or invalid"
}

解决方法:

检查 API Key 是否正确复制
确认 Authorization header 格式: Bearer YOUR_API_KEY
访问 API Key 管理页面重新生成

问题 2: 任务一直处于 waiting 状态

原因:

服务器负载高
任务排队中

解决方法:

延长轮询等待时间（最多 5 分钟）
使用 callback URL 代替轮询
选择 Fast/Turbo 速度（Midjourney）

问题 3: 生成的图片质量不理想

优化方法:

优化 Prompt: - 增加更多细节描述 - 指定风格和质量关键词（photorealistic, 4K, professional 等） - 参考优秀 Prompt 示例
调整参数:

json { "stylization": 200, // 增加艺术感 "variety": 10, // 保持一致性 "version": "7" // 使用最新版本 }

切换模型: - Nano Banana 不满意 → 试试 Midjourney - 需要更艺术化 → 提高 stylization 值

问题 4: 积分不足

错误信息:

{
  "code": 402,
  "msg": "Insufficient Credits"
}

解决方法:

登录 Kie.ai 充值积分
优化使用策略：优先使用 Nano Banana 节省积分
批量生成时控制数量

问题 5: 图片 URL 过期

说明:

Nano Banana 图片保留 15 天
Midjourney 图片保留 15 天

解决方法:

// 生成后立即下载保存到本地
const imageUrl = await generateImage(prompt);
await downloadImage(imageUrl, 'blog-cover.png');

📚 相关资源

API 文档: Kie.ai API Docs
API Key 管理: https://kie.ai/api-key
Prompt 参考: Midjourney Prompt Guide
本项目配图指南: GENERATE-IMAGES-GUIDE.md

🎯 快速命令参考

Nano Banana - Text-to-Image

curl -X POST https://api.kie.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $KIE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/nano-banana",
    "input": {
      "prompt": "YOUR_PROMPT_HERE",
      "image_size": "16:9",
      "output_format": "png"
    }
  }'

Midjourney - Text-to-Image

curl -X POST https://api.kie.ai/api/v1/mj/generate \
  -H "Authorization: Bearer $KIE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskType": "mj_txt2img",
    "prompt": "YOUR_PROMPT_HERE --ar 16:9 --v 7",
    "speed": "relaxed",
    "aspectRatio": "16:9",
    "version": "7"
  }'

查询任务状态

# Nano Banana
curl "https://api.kie.ai/api/v1/jobs/recordInfo?taskId=YOUR_TASK_ID" \
  -H "Authorization: Bearer $KIE_API_KEY"

# Midjourney
curl "https://api.kie.ai/api/v1/mj/record-info?taskId=YOUR_TASK_ID" \
  -H "Authorization: Bearer $KIE_API_KEY"

版本: 1.0.0 更新日期: 2026-01-23 维护: SEO Team

博客特色图片生成操作指南

目的: 为 8 篇缺少特色图片的博客文章生成高质量 featured images

预计时间: 每篇 7-10 分钟，总计 ~60 分钟 预计成本: 1,120 credits (~$11.20)

📋 准备工作

1. 确认账户积分

登录 https://seedvr2.net/settings/credits
确保至少有 1,120 credits 可用
如果不足，请先充值

2. 准备工具

图片压缩工具: TinyPNG 或 ImageOptim
文本编辑器: VSCode 或任意编辑器
浏览器: Chrome/Safari（用于下载图片）

🎨 逐篇生成流程

所有 prompt 已保存在 scripts/generate-featured-images.json，下面是快速操作步骤：

文章 1: Ultimate Upscaling Guide (对比类)

文件: tutorials-seedvr2-ultimate-upscaling-guide-2026.mdx

步骤:

打开 https://seedvr2.net/models/ideogram
复制以下 Prompt: Split composition with two contrasting AI neural network visualizations, left side in electric blue tones showing SeedVR2 architecture, right side in vibrant purple tones showing FlashVSR structure, clean separation line with subtle glow, abstract data streams and neural pathways flowing between them, modern tech aesthetic, professional 3D rendering, 16:9 landscape format, high quality, photorealistic style
Negative Prompt: text, labels, words, letters, watermark, logo, low quality, blurry, cartoon, anime, people, faces
设置: - Style: DESIGN - Image Size: landscape_16_9 - Speed: BALANCED - Num Images: 2
点击 "Generate" (消耗 140 credits)
等待 ~30 秒生成完成
从 2 张中选择最佳的一张
下载图片，命名为: tutorials-seedvr2-ultimate-upscaling-guide-2026-featured.jpg
使用 TinyPNG 压缩到 <200KB
保存到: public/images/blog/tutorials/
更新文章 frontmatter: yaml published: true image: '/images/blog/tutorials/tutorials-seedvr2-ultimate-upscaling-guide-2026-featured.jpg' imageAlt: 'Visual comparison of SeedVR2 and FlashVSR AI upscaling architectures with neural network visualization'

文章 2: Low VRAM Workflow (教程类)

文件: tutorials-seedvr2-low-vram-workflow-2026.mdx

步骤:

打开 Ideogram 页面
复制 Prompt: A powerful modern GPU graphics card with limited memory chips glowing in soft blue light, abstract memory optimization visualization flowing around it with green data streams, clean tech laboratory background, professional studio lighting with soft blue and green accents, photorealistic 3D render, shallow depth of field, high detail, 16:9 aspect ratio, ultra realistic
Negative Prompt: text, watermark, logo, typography, words, letters, low quality, blurry, cartoon, people, faces, hands
设置: - Style: REALISTIC - Image Size: landscape_16_9 - Speed: BALANCED - Num Images: 2
Generate (140 credits)
下载最佳图片: tutorials-seedvr2-low-vram-workflow-2026-featured.jpg
压缩 <200KB
保存到 public/images/blog/tutorials/
更新 frontmatter: yaml published: true image: '/images/blog/tutorials/tutorials-seedvr2-low-vram-workflow-2026-featured.jpg' imageAlt: 'Modern GPU with memory optimization visualization for SeedVR2 low VRAM workflow on 8GB graphics cards'

文章 3: ComfyUI Low VRAM Guide (教程类)

文件: tutorials-seedvr2-comfyui-low-vram-guide-2026.mdx

Prompt:

Abstract ComfyUI node graph interface with glowing connections, GPU memory tiles floating in organized grid pattern, soft teal and blue lighting, clean minimalist tech background, professional 3D visualization, data flowing through interconnected nodes, 16:9 landscape format, high quality rendering, modern tech aesthetic

Negative Prompt:

text, UI elements, buttons, labels, watermark, logo, low quality, blurry, cartoon, people

设置: REALISTIC, landscape_16_9, BALANCED, 2 images

下载命名: tutorials-seedvr2-comfyui-low-vram-guide-2026-featured.jpg

Alt Text:

ComfyUI node workflow visualization for SeedVR2 tiled upscaling with memory management on 8-12GB GPUs

文章 4: Top Free Upscaler (评测类)

文件: tutorials-seedvr2-top-free-upscaler-2026.mdx

Prompt:

Premium AI upscaling trophy or award concept in spotlight setup, glowing golden and blue neural network visualization surrounding it, dramatic lighting with soft blue and purple accents, floating in clean void, high-end product photography style, professional studio lighting with subtle reflections, 16:9 aspect ratio, ultra detailed, photorealistic rendering

Negative Prompt:

text, numbers, words, ratings, stars, watermark, logo, low quality, blurry, cartoon

设置: REALISTIC, landscape_16_9, BALANCED, 2 images

下载命名: tutorials-seedvr2-top-free-upscaler-2026-featured.jpg

Alt Text:

Premium visualization representing SeedVR2 as the top free AI video upscaler with quality assessment

文章 5: Video Upscaling in ComfyUI (教程类)

文件: tutorials-seedvr2-video-upscaling-comfyui-2026.mdx

Prompt:

Abstract video frames flowing through AI processing pipeline, glowing blue resolution enhancement visualization with upward arrows, clean ComfyUI-style node connections, professional tech aesthetic, soft blue and green lighting, 16:9 landscape format, photorealistic 3D render, depth of field effect, high quality

Negative Prompt:

text, watermark, logo, UI elements, buttons, words, low quality, blurry, cartoon, people

设置: REALISTIC, landscape_16_9, BALANCED, 2 images

下载命名: tutorials-seedvr2-video-upscaling-comfyui-2026-featured.jpg

Alt Text:

Video processing pipeline visualization for SeedVR2 batch upscaling workflow in ComfyUI

文章 6: 8GB SRPO Secrets (教程类)

文件: tutorials-seedvr2-8gb-srpo-secrets-2026.mdx

Prompt:

A sleek GPU memory chip revealing intricate glowing circuits underneath in layers, mysterious soft blue and purple lighting with golden accents for premium feel, tech laboratory environment, depth of field effect focusing on circuit details, professional macro photography style, futuristic aesthetic, 16:9 landscape format, ultra detailed photorealistic rendering

Negative Prompt:

text, numbers, labels, watermark, logo, cartoon, low quality, blurry, people, hands

设置: REALISTIC, landscape_16_9, BALANCED, 2 images

下载命名: tutorials-seedvr2-8gb-srpo-secrets-2026-featured.jpg

Alt Text:

Detailed GPU circuit visualization revealing SeedVR2 and SRPO optimization secrets for 8GB graphics cards

文章 7: One-Click Upscaler (教程类)

文件: tutorials-seedvr2-one-click-upscaler-2026.mdx

Prompt:

A single large glowing blue button surrounded by abstract upward-flowing light particles and resolution enhancement sparkles, clean minimalist void background with soft gradient, soft ambient lighting creating magical feel, professional product photography style, shallow depth of field, 16:9 aspect ratio, photorealistic rendering, beginner-friendly aesthetic

Negative Prompt:

text, words, labels, typography, watermark, logo, people, hands, complex elements, low quality, blurry

设置: REALISTIC, landscape_16_9, BALANCED, 2 images

下载命名: tutorials-seedvr2-one-click-upscaler-2026-featured.jpg

Alt Text:

Simple one-click button visualization for beginner-friendly SeedVR2 upscaling workflow in ComfyUI

文章 8: Complete Guide (指南类)

文件: tutorials-seedvr2-complete-guide-2026.mdx

Prompt:

An abstract visualization of a learning pathway with glowing orange and yellow guideposts, futuristic navigation interface elements floating in space, clean geometric shapes forming a journey from beginner to expert, soft warm lighting creating welcoming atmosphere, depth and perspective showing progress, professional 3D rendering, 16:9 landscape format, modern educational aesthetic

Negative Prompt:

text, labels, words, arrows, numbers, watermark, logo, people, complex UI, low quality, blurry

设置: DESIGN, landscape_16_9, BALANCED, 2 images

下载命名: tutorials-seedvr2-complete-guide-2026-featured.jpg

Alt Text:

Learning pathway visualization for SeedVR2 complete beginner's guide to 4K video upscaling

🔄 批量更新 Frontmatter 脚本

生成并下载所有图片后，使用此脚本批量更新 frontmatter:

#!/bin/bash
# scripts/update-image-frontmatter.sh

# 定义图片和 Alt 文本映射
declare -A images
images["tutorials-seedvr2-ultimate-upscaling-guide-2026"]="Visual comparison of SeedVR2 and FlashVSR AI upscaling architectures with neural network visualization"
images["tutorials-seedvr2-low-vram-workflow-2026"]="Modern GPU with memory optimization visualization for SeedVR2 low VRAM workflow on 8GB graphics cards"
images["tutorials-seedvr2-comfyui-low-vram-guide-2026"]="ComfyUI node workflow visualization for SeedVR2 tiled upscaling with memory management on 8-12GB GPUs"
images["tutorials-seedvr2-top-free-upscaler-2026"]="Premium visualization representing SeedVR2 as the top free AI video upscaler with quality assessment"
images["tutorials-seedvr2-video-upscaling-comfyui-2026"]="Video processing pipeline visualization for SeedVR2 batch upscaling workflow in ComfyUI"
images["tutorials-seedvr2-8gb-srpo-secrets-2026"]="Detailed GPU circuit visualization revealing SeedVR2 and SRPO optimization secrets for 8GB graphics cards"
images["tutorials-seedvr2-one-click-upscaler-2026"]="Simple one-click button visualization for beginner-friendly SeedVR2 upscaling workflow in ComfyUI"
images["tutorials-seedvr2-complete-guide-2026"]="Learning pathway visualization for SeedVR2 complete beginner's guide to 4K video upscaling"

# 遍历更新
for slug in "${!images[@]}"; do
  file="content/blog/tutorials/${slug}.mdx"
  alt="${images[$slug]}"

  if [ -f "$file" ]; then
    # 检查是否已有 image 字段
    if grep -q "^image:" "$file"; then
      echo "⏭️  跳过（已有图片）: $slug"
    else
      # 在 published: true 后插入 image 和 imageAlt
      sed -i '' "/published: true/a\\
image: \"/images/blog/tutorials/${slug}-featured.jpg\"\\
imageAlt: \"$alt\"
" "$file"
      echo "✅ 已更新: $slug"
    fi
  else
    echo "❌ 文件不存在: $file"
  fi
done

echo ""
echo "🎉 批量更新完成！"

使用方法:

chmod +x scripts/update-image-frontmatter.sh
./scripts/update-image-frontmatter.sh

✅ 质量检查清单

生成每张图片后，确认:

[ ] 图片尺寸为 16:9（推荐 1920×1080 或 1600×900）
[ ] 文件大小 <200KB（压缩后）
[ ] 文件格式为 JPG（质量 85%）
[ ] 文件命名正确: [slug]-featured.jpg
[ ] 保存到 public/images/blog/tutorials/
[ ] 无文字、水印、logo
[ ] 视觉清晰，主题相关
[ ] 色彩协调，符合文章类型
[ ] Frontmatter 已更新 image: 和 imageAlt:
[ ] 本地预览正常显示

📊 进度追踪

#	文章	图片生成	下载优化	Frontmatter	完成
1	ultimate-upscaling-guide	⬜	⬜	⬜	⬜
2	low-vram-workflow	⬜	⬜	⬜	⬜
3	comfyui-low-vram-guide	⬜	⬜	⬜	⬜
4	top-free-upscaler	⬜	⬜	⬜	⬜
5	video-upscaling-comfyui	⬜	⬜	⬜	⬜
6	8gb-srpo-secrets	⬜	⬜	⬜	⬜
7	one-click-upscaler	⬜	⬜	⬜	⬜
8	complete-guide	⬜	⬜	⬜	⬜

总进度: 0/8 (0%)

🚀 完成后

Git 提交:

```bash git add public/images/blog/tutorials/.jpg git add content/blog/tutorials/.mdx git commit -m "feat: add featured images for 8 blog articles

Generate high-quality images using Ideogram API
All images optimized <200KB
Update frontmatter with image paths and alt text
Total cost: 1,120 credits (~$11.20)

Co-Authored-By: Claude Sonnet 4.5 noreply@anthropic.com"

git push origin main ```

验证部署: - 等待 CI/CD 构建完成 - 访问 https://seedvr2.net/blog?category=tutorials - 确认所有 15 篇文章都显示特色图片
更新文档: - 更新 SEO/seoskill.md 统计数据 - 标记任务完成: ✅ 15/15 文章已配图

预计完成时间: 60-80 分钟 预计成本: 1,120 credits (~$11.20) 预期效果: 博客视觉质量显著提升，SEO 图片搜索排名改善

开始时间: *_ 完成时间: _ 实际成本: _* credits

本文档为站内渲染。原始文件本地路径：saas/source/seo-llm/raw-seo知识库-seo方法论-seo-knowledge-base-Guideline-AI-IMAGE-GENE-a13004.md（仅本地保留，不入库不部署）