amz-pic-flow/test-vision-vs-manual.js

/**
 * 对比测试：Vision自动提取 vs 手写描述
 * 生成Main_02平铺图，对比两者效果
 */

require('dotenv').config();
const axios = require('axios');
const fs = require('fs');
const path = require('path');
const { S3Client, PutObjectCommand } = require('@aws-sdk/client-s3');

const API_KEY = 'G9rXx3Ag2Xfa7Gs8zou6t6HqeZ';
const API_BASE = 'https://api.wuyinkeji.com/api';
const OUTPUT_DIR = path.join(__dirname, 'output_comparison');
const MATERIAL_DIR = path.join(__dirname, '素材/素材/已有的素材');

// R2客户端
const r2Client = new S3Client({
  region: 'auto',
  endpoint: `https://${process.env.R2_ACCOUNT_ID}.r2.cloudflarestorage.com`,
  credentials: {
    accessKeyId: process.env.R2_ACCESS_KEY_ID,
    secretAccessKey: process.env.R2_SECRET_ACCESS_KEY,
  },
  forcePathStyle: true,
});

async function uploadToR2(filePath) {
  const fileName = `compare-${Date.now()}-${path.basename(filePath)}`;
  const fileBuffer = fs.readFileSync(filePath);
  
  await r2Client.send(new PutObjectCommand({
    Bucket: process.env.R2_BUCKET_NAME || 'ai-flow',
    Key: fileName,
    Body: fileBuffer,
    ContentType: filePath.endsWith('.png') ? 'image/png' : 'image/jpeg',
  }));

  return process.env.R2_PUBLIC_DOMAIN
    ? `${process.env.R2_PUBLIC_DOMAIN}/${fileName}`
    : `https://${process.env.R2_ACCOUNT_ID}.r2.cloudflarestorage.com/${process.env.R2_BUCKET_NAME}/${fileName}`;
}

// ========================================
// 方案A: 手写的Golden Description (V2版)
// ========================================
const MANUAL_GOLDEN_DESC = `
EXACT PRODUCT APPEARANCE (MUST MATCH PRECISELY):
- Shape: 8-PETAL FLOWER/FAN shape, C-shaped opening (like Pac-Man), NOT a full circle
- Color: ICE BLUE / Light aqua blue (#C5E8ED approximate)
- Material: Glossy waterproof PU fabric with visible stitching lines between petals
- Edge binding: TEAL/TURQUOISE color binding around the inner neck hole
- Closure: White velcro strap on one petal end
- Logo: "TOUCHDOG®" embroidered in matching blue thread on one petal (small, subtle)
- Texture: Smooth, slightly shiny, NOT matte, NOT cotton fabric
- Structure: Soft but structured, maintains petal shape, NOT floppy

CRITICAL PROHIBITIONS:
- ❌ NO printed patterns or colorful fabric designs
- ❌ NO hard plastic transparent cones
- ❌ NO fully circular/closed shapes
- ❌ NO matte cotton or fleece textures
- ❌ NO random brand logos or text
`;

// ========================================
// 方案B: Vision自动提取的Golden Description
// ========================================
const VISION_GOLDEN_DESC = `
EXACT PRODUCT APPEARANCE (AUTO-EXTRACTED BY VISION):
- Shape: 7-PETAL FLOWER/FAN shape, C-shaped opening
- Color: PASTEL ICE BLUE (#C3E6E8)
- Material: soft matte/satin synthetic fabric (likely water-resistant polyester/nylon) with smooth, padded/quilted panels texture
- Edge binding: Mint Green (slightly more saturated than body) ribbed knit elastic fabric around inner neck hole
- Closure: white velcro (hook and loop) on large rectangular strip covering the entire bottom-left terminal segment
- Logo: "TOUCHDOG®" embroidered on centered on the middle-right segment

UNIQUE FEATURES:
- Scalloped outer edge resembling flower petals
- Soft ribbed knit neckline for comfort
- Radial stitching lines creating distinct padded zones
- Large surface area velcro for adjustable sizing

CRITICAL PROHIBITIONS:
- ❌ NO printed patterns or colorful fabric designs
- ❌ NO hard plastic transparent cones
- ❌ NO fully circular/closed shapes (must have C-opening)
- ❌ NO random brand logos or text
`;

// Main_02 平铺图 Prompt模板
function createMain02Prompt(goldenDesc) {
  return `
[PRODUCT FLAT LAY PHOTO - WHITE BACKGROUND]

${goldenDesc}

SCENE REQUIREMENTS:
- Product flat lay on PURE WHITE background (#FFFFFF)
- Shot from directly above (bird's eye view / top-down angle)
- Show the full C-shaped opening clearly (gap between velcro ends visible)
- All petals/segments fully visible and spread out
- Clean studio lighting, MINIMAL shadows
- Product fills 70-80% of frame
- Professional Amazon product photography style

OUTPUT: High quality 1:1 aspect ratio product photo, 8K resolution
`.trim();
}

// 生图任务提交
async function submitImageTask(prompt, refImageUrl) {
  const payload = {
    key: API_KEY,
    prompt: prompt,
    img_url: refImageUrl,
    aspectRatio: '1:1',
    imageSize: '1K'
  };

  const response = await axios.post(`${API_BASE}/img/nanoBanana-pro`, payload);
  const taskId = response.data.data?.id;
  
  if (!taskId) {
    throw new Error('No task ID returned');
  }
  return taskId;
}

// 轮询图片结果
async function pollImageResult(taskId) {
  let attempts = 0;
  const maxAttempts = 60;
  
  while (attempts < maxAttempts) {
    const response = await axios.get(`${API_BASE}/img/drawDetail`, {
      params: { key: API_KEY, id: taskId }
    });
    
    const data = response.data.data;
    
    if (data && data.status === 2 && data.image_url) {
      return data.image_url;
    } else if (data && data.status === 3) {
      throw new Error('Generation failed: ' + (data.fail_reason || 'Unknown'));
    }
    
    process.stdout.write('.');
    await new Promise(r => setTimeout(r, 2000));
    attempts++;
  }
  throw new Error('Timeout waiting for image');
}

// 生成单张图
async function generateImage(name, prompt, refUrl) {
  console.log(`\n🎨 生成: ${name}`);
  console.log('   提交任务...');
  
  const taskId = await submitImageTask(prompt, refUrl);
  console.log(`   Task ID: ${taskId}`);
  
  const imageUrl = await pollImageResult(taskId);
  console.log('\n   ✓ 生成成功');
  
  // 下载保存
  const imgRes = await axios.get(imageUrl, { responseType: 'arraybuffer' });
  const outputPath = path.join(OUTPUT_DIR, `${name}.jpg`);
  fs.writeFileSync(outputPath, imgRes.data);
  console.log(`   ✓ 保存: ${outputPath}`);
  
  return outputPath;
}

async function main() {
  if (!fs.existsSync(OUTPUT_DIR)) fs.mkdirSync(OUTPUT_DIR, { recursive: true });

  console.log('='.repeat(60));
  console.log('🔬 Vision提取 vs 手写描述 对比测试');
  console.log('='.repeat(60));

  // 上传参考图
  console.log('\n📤 上传参考图...');
  const flatImgPath = path.join(MATERIAL_DIR, 'IMG_5683.png');
  const refUrl = await uploadToR2(flatImgPath);
  console.log('   参考图URL:', refUrl);

  // 生成两个版本的Main_02
  const results = [];

  // 版本A: 手写描述
  console.log('\n' + '='.repeat(60));
  console.log('📝 版本A: 手写Golden Description');
  console.log('='.repeat(60));
  const promptA = createMain02Prompt(MANUAL_GOLDEN_DESC);
  console.log('Prompt预览:\n', promptA.substring(0, 500) + '...');
  
  try {
    const pathA = await generateImage('Main_02_Manual', promptA, refUrl);
    results.push({ name: 'Manual', path: pathA, success: true });
  } catch (e) {
    console.error('版本A失败:', e.message);
    results.push({ name: 'Manual', success: false, error: e.message });
  }

  // 版本B: Vision提取
  console.log('\n' + '='.repeat(60));
  console.log('🤖 版本B: Vision自动提取Golden Description');
  console.log('='.repeat(60));
  const promptB = createMain02Prompt(VISION_GOLDEN_DESC);
  console.log('Prompt预览:\n', promptB.substring(0, 500) + '...');
  
  try {
    const pathB = await generateImage('Main_02_Vision', promptB, refUrl);
    results.push({ name: 'Vision', path: pathB, success: true });
  } catch (e) {
    console.error('版本B失败:', e.message);
    results.push({ name: 'Vision', success: false, error: e.message });
  }

  // 总结
  console.log('\n' + '='.repeat(60));
  console.log('📊 对比结果');
  console.log('='.repeat(60));
  console.log(`输出目录: ${OUTPUT_DIR}`);
  console.log('\n生成结果:');
  results.forEach(r => {
    if (r.success) {
      console.log(`  ✅ ${r.name}: ${r.path}`);
    } else {
      console.log(`  ❌ ${r.name}: ${r.error}`);
    }
  });
  
  console.log('\n💡 请手动对比两张图片，评估Vision自动提取的效果是否达标');
}

main().catch(console.error);