diff --git a/src/app/actions/scan-label.ts b/src/app/actions/scan-label.ts
index 6dfa9d4..60aac16 100644
--- a/src/app/actions/scan-label.ts
+++ b/src/app/actions/scan-label.ts
@@ -6,6 +6,7 @@ import { createClient } from '@/lib/supabase/server';
 import { createHash } from 'crypto';
 import { trackApiUsage } from '@/services/track-api-usage';
 import { checkCreditBalance, deductCredits } from '@/services/credit-service';
+import sharp from 'sharp';
 
 // Native Schema Definition for Gemini API
 const metadataSchema = {
@@ -100,14 +101,23 @@ export async function scanLabel(input: any): Promise<AnalysisResponse> {
             };
         }
 
-        // Step 3: Base64 Encoding
+        // Step 3: AI-Specific Image Optimization (Grayscale, Normalize, Resize)
+        const startOptimization = performance.now();
+        const optimizedBuffer = await sharp(buffer)
+            .resize(1024, 1024, { fit: 'inside', withoutEnlargement: true })
+            .grayscale()
+            .normalize()
+            .toBuffer();
+        const endOptimization = performance.now();
+
+        // Step 4: Base64 Encoding
         const startEncoding = performance.now();
-        const base64Data = buffer.toString('base64');
-        const mimeType = file.type || 'image/webp';
-        const uploadSize = buffer.length;
+        const base64Data = optimizedBuffer.toString('base64');
+        const mimeType = 'image/webp';
+        const uploadSize = optimizedBuffer.length;
         const endEncoding = performance.now();
 
-        // Step 4: Model Initialization & Step 5: API Call
+        // Step 5: Model Initialization & Step 6: API Call
         const startAiTotal = performance.now();
         let jsonData;
         let validatedData;
@@ -164,6 +174,7 @@ export async function scanLabel(input: any): Promise<AnalysisResponse> {
                 data: validatedData,
                 perf: {
                     imagePrep: endImagePrep - startImagePrep,
+                    optimization: endOptimization - startOptimization,
                     cacheCheck: endCacheCheck - startCacheCheck,
                     encoding: endEncoding - startEncoding,
                     modelInit: endModelInit - startModelInit,