diff --git a/src/app/api/generate/providers/gemini.ts b/src/app/api/generate/providers/gemini.ts
new file mode 100644
index 00000000..aa2a3bf8
--- /dev/null
+++ b/src/app/api/generate/providers/gemini.ts
@@ -0,0 +1,188 @@
+/**
+ * Gemini Provider for Generate API Route
+ *
+ * Handles image generation using Google's Gemini API models.
+ */
+
+import { NextResponse } from "next/server";
+import { GoogleGenAI } from "@google/genai";
+import { GenerateResponse, ModelType } from "@/types";
+
+/**
+ * Map model types to Gemini model IDs
+ */
+export const MODEL_MAP: Record<ModelType, string> = {
+  "nano-banana": "gemini-2.5-flash-image",
+  "nano-banana-pro": "gemini-3-pro-image-preview",
+};
+
+/**
+ * Generate image using Gemini API (legacy/default path)
+ */
+export async function generateWithGemini(
+  requestId: string,
+  apiKey: string,
+  prompt: string,
+  images: string[],
+  model: ModelType,
+  aspectRatio?: string,
+  resolution?: string,
+  useGoogleSearch?: boolean
+): Promise<NextResponse<GenerateResponse>> {
+  console.log(`[API:${requestId}] Gemini generation - Model: ${model}, Images: ${images?.length || 0}, Prompt: ${prompt?.length || 0} chars`);
+
+  // Extract base64 data and MIME types from data URLs
+  const imageData = (images || []).map((image, idx) => {
+    if (image.includes("base64,")) {
+      const [header, data] = image.split("base64,");
+      // Extract MIME type from header (e.g., "data:image/png;" -> "image/png")
+      const mimeMatch = header.match(/data:([^;]+)/);
+      const mimeType = mimeMatch ? mimeMatch[1] : "image/png";
+      console.log(`[API:${requestId}]   Image ${idx + 1}: ${mimeType}, ${(data.length / 1024).toFixed(1)}KB`);
+      return { data, mimeType };
+    }
+    console.log(`[API:${requestId}]   Image ${idx + 1}: raw, ${(image.length / 1024).toFixed(1)}KB`);
+    return { data: image, mimeType: "image/png" };
+  });
+
+  // Initialize Gemini client
+  const ai = new GoogleGenAI({ apiKey });
+
+  // Build request parts array with prompt and all images
+  const requestParts: Array<{ text: string } | { inlineData: { mimeType: string; data: string } }> = [
+    { text: prompt },
+    ...imageData.map(({ data, mimeType }) => ({
+      inlineData: {
+        mimeType,
+        data,
+      },
+    })),
+  ];
+
+  // Build config object based on model capabilities
+  const config: Record<string, unknown> = {
+    responseModalities: ["IMAGE", "TEXT"],
+  };
+
+  // Add imageConfig for both models (both support aspect ratio)
+  if (aspectRatio) {
+    config.imageConfig = {
+      aspectRatio,
+    };
+  }
+
+  // Add resolution only for Nano Banana Pro
+  if (model === "nano-banana-pro" && resolution) {
+    if (!config.imageConfig) {
+      config.imageConfig = {};
+    }
+    (config.imageConfig as Record<string, unknown>).imageSize = resolution;
+  }
+
+  // Add tools array for Google Search (only Nano Banana Pro)
+  const tools = [];
+  if (model === "nano-banana-pro" && useGoogleSearch) {
+    tools.push({ googleSearch: {} });
+  }
+
+  console.log(`[API:${requestId}] Config: ${JSON.stringify(config)}`);
+
+  // Make request to Gemini
+  const geminiStartTime = Date.now();
+
+  const response = await ai.models.generateContent({
+    model: MODEL_MAP[model],
+    contents: [
+      {
+        role: "user",
+        parts: requestParts,
+      },
+    ],
+    config,
+    ...(tools.length > 0 && { tools }),
+  });
+
+  const geminiDuration = Date.now() - geminiStartTime;
+  console.log(`[API:${requestId}] Gemini API completed in ${geminiDuration}ms`);
+
+  // Extract image from response
+  const candidates = response.candidates;
+
+  if (!candidates || candidates.length === 0) {
+    console.error(`[API:${requestId}] No candidates in Gemini response`);
+    return NextResponse.json<GenerateResponse>(
+      {
+        success: false,
+        error: "No response from AI model",
+      },
+      { status: 500 }
+    );
+  }
+
+  const parts = candidates[0].content?.parts;
+  console.log(`[API:${requestId}] Response parts: ${parts?.length || 0}`);
+
+  if (!parts) {
+    console.error(`[API:${requestId}] No parts in Gemini candidate content`);
+    return NextResponse.json<GenerateResponse>(
+      {
+        success: false,
+        error: "No content in response",
+      },
+      { status: 500 }
+    );
+  }
+
+  // Find image part in response
+  for (const part of parts) {
+    if (part.inlineData && part.inlineData.data) {
+      const mimeType = part.inlineData.mimeType || "image/png";
+      const imgData = part.inlineData.data;
+      const imageSizeKB = (imgData.length / 1024).toFixed(1);
+
+      console.log(`[API:${requestId}] Output image: ${mimeType}, ${imageSizeKB}KB`);
+
+      const dataUrl = `data:${mimeType};base64,${imgData}`;
+
+      const responsePayload = { success: true, image: dataUrl };
+      const responseSize = JSON.stringify(responsePayload).length;
+      const responseSizeMB = (responseSize / (1024 * 1024)).toFixed(2);
+
+      if (responseSize > 4.5 * 1024 * 1024) {
+        console.warn(`[API:${requestId}] Response size (${responseSizeMB}MB) approaching Next.js 5MB limit`);
+      }
+
+      console.log(`[API:${requestId}] SUCCESS - Returning ${responseSizeMB}MB payload`);
+
+      // Create response with explicit headers to handle large payloads
+      const resp = NextResponse.json<GenerateResponse>(responsePayload);
+      resp.headers.set('Content-Type', 'application/json');
+      resp.headers.set('Content-Length', responseSize.toString());
+
+      return resp;
+    }
+  }
+
+  // If no image found, check for text error
+  for (const part of parts) {
+    if (part.text) {
+      console.error(`[API:${requestId}] Gemini returned text instead of image: ${part.text.substring(0, 100)}`);
+      return NextResponse.json<GenerateResponse>(
+        {
+          success: false,
+          error: `Model returned text instead of image: ${part.text.substring(0, 200)}`,
+        },
+        { status: 500 }
+      );
+    }
+  }
+
+  console.error(`[API:${requestId}] No image or text found in Gemini response`);
+  return NextResponse.json<GenerateResponse>(
+    {
+      success: false,
+      error: "No image in response",
+    },
+    { status: 500 }
+  );
+}
diff --git a/src/app/api/generate/route.ts b/src/app/api/generate/route.ts
index 122e03cf..5d9a0a4c 100644
--- a/src/app/api/generate/route.ts
+++ b/src/app/api/generate/route.ts
@@ -11,10 +11,8 @@
  * Images are uploaded to fal CDN before submission to avoid payload size issues.
  */
 import { NextRequest, NextResponse } from "next/server";
-import { GoogleGenAI } from "@google/genai";
 import { GenerateRequest, GenerateResponse, ModelType, SelectedModel, ProviderType } from "@/types";
-import { GenerationInput, GenerationOutput, ProviderModel } from "@/lib/providers/types";
-import { uploadImageForUrl, shouldUseImageUrl, deleteImages } from "@/lib/images";
+import { GenerationInput, GenerationOutput } from "@/lib/providers/types";
 import { validateMediaUrl } from "@/utils/urlValidation";
 import {
   INPUT_PATTERNS,
@@ -24,15 +22,11 @@ import {
   coerceParameterTypes,
   getInputMappingFromSchema,
 } from "./schemaUtils";
+import { generateWithGemini } from "./providers/gemini";
 
 export const maxDuration = 300; // 5 minute timeout (Vercel hobby plan limit)
 export const dynamic = 'force-dynamic'; // Ensure this route is always dynamic
 
-// Map model types to Gemini model IDs
-const MODEL_MAP: Record<ModelType, string> = {
-  "nano-banana": "gemini-2.5-flash-image", // Updated to correct model name
-  "nano-banana-pro": "gemini-3-pro-image-preview",
-};
 
 /**
  * Extended request format that supports both legacy and multi-provider requests
@@ -44,176 +38,6 @@ interface MultiProviderGenerateRequest extends GenerateRequest {
   dynamicInputs?: Record<string, string | string[]>;
 }
 
-/**
- * Generate image using Gemini API (legacy/default path)
- */
-async function generateWithGemini(
-  requestId: string,
-  apiKey: string,
-  prompt: string,
-  images: string[],
-  model: ModelType,
-  aspectRatio?: string,
-  resolution?: string,
-  useGoogleSearch?: boolean
-): Promise<NextResponse<GenerateResponse>> {
-  console.log(`[API:${requestId}] Gemini generation - Model: ${model}, Images: ${images?.length || 0}, Prompt: ${prompt?.length || 0} chars`);
-
-  // Extract base64 data and MIME types from data URLs
-  const imageData = (images || []).map((image, idx) => {
-    if (image.includes("base64,")) {
-      const [header, data] = image.split("base64,");
-      // Extract MIME type from header (e.g., "data:image/png;" -> "image/png")
-      const mimeMatch = header.match(/data:([^;]+)/);
-      const mimeType = mimeMatch ? mimeMatch[1] : "image/png";
-      console.log(`[API:${requestId}]   Image ${idx + 1}: ${mimeType}, ${(data.length / 1024).toFixed(1)}KB`);
-      return { data, mimeType };
-    }
-    console.log(`[API:${requestId}]   Image ${idx + 1}: raw, ${(image.length / 1024).toFixed(1)}KB`);
-    return { data: image, mimeType: "image/png" };
-  });
-
-  // Initialize Gemini client
-  const ai = new GoogleGenAI({ apiKey });
-
-  // Build request parts array with prompt and all images
-  const requestParts: Array<{ text: string } | { inlineData: { mimeType: string; data: string } }> = [
-    { text: prompt },
-    ...imageData.map(({ data, mimeType }) => ({
-      inlineData: {
-        mimeType,
-        data,
-      },
-    })),
-  ];
-
-  // Build config object based on model capabilities
-  const config: Record<string, unknown> = {
-    responseModalities: ["IMAGE", "TEXT"],
-  };
-
-  // Add imageConfig for both models (both support aspect ratio)
-  if (aspectRatio) {
-    config.imageConfig = {
-      aspectRatio,
-    };
-  }
-
-  // Add resolution only for Nano Banana Pro
-  if (model === "nano-banana-pro" && resolution) {
-    if (!config.imageConfig) {
-      config.imageConfig = {};
-    }
-    (config.imageConfig as Record<string, unknown>).imageSize = resolution;
-  }
-
-  // Add tools array for Google Search (only Nano Banana Pro)
-  const tools = [];
-  if (model === "nano-banana-pro" && useGoogleSearch) {
-    tools.push({ googleSearch: {} });
-  }
-
-  console.log(`[API:${requestId}] Config: ${JSON.stringify(config)}`);
-
-  // Make request to Gemini
-  const geminiStartTime = Date.now();
-
-  const response = await ai.models.generateContent({
-    model: MODEL_MAP[model],
-    contents: [
-      {
-        role: "user",
-        parts: requestParts,
-      },
-    ],
-    config,
-    ...(tools.length > 0 && { tools }),
-  });
-
-  const geminiDuration = Date.now() - geminiStartTime;
-  console.log(`[API:${requestId}] Gemini API completed in ${geminiDuration}ms`);
-
-  // Extract image from response
-  const candidates = response.candidates;
-
-  if (!candidates || candidates.length === 0) {
-    console.error(`[API:${requestId}] No candidates in Gemini response`);
-    return NextResponse.json<GenerateResponse>(
-      {
-        success: false,
-        error: "No response from AI model",
-      },
-      { status: 500 }
-    );
-  }
-
-  const parts = candidates[0].content?.parts;
-  console.log(`[API:${requestId}] Response parts: ${parts?.length || 0}`);
-
-  if (!parts) {
-    console.error(`[API:${requestId}] No parts in Gemini candidate content`);
-    return NextResponse.json<GenerateResponse>(
-      {
-        success: false,
-        error: "No content in response",
-      },
-      { status: 500 }
-    );
-  }
-
-  // Find image part in response
-  for (const part of parts) {
-    if (part.inlineData && part.inlineData.data) {
-      const mimeType = part.inlineData.mimeType || "image/png";
-      const imgData = part.inlineData.data;
-      const imageSizeKB = (imgData.length / 1024).toFixed(1);
-
-      console.log(`[API:${requestId}] Output image: ${mimeType}, ${imageSizeKB}KB`);
-
-      const dataUrl = `data:${mimeType};base64,${imgData}`;
-
-      const responsePayload = { success: true, image: dataUrl };
-      const responseSize = JSON.stringify(responsePayload).length;
-      const responseSizeMB = (responseSize / (1024 * 1024)).toFixed(2);
-
-      if (responseSize > 4.5 * 1024 * 1024) {
-        console.warn(`[API:${requestId}] Response size (${responseSizeMB}MB) approaching Next.js 5MB limit`);
-      }
-
-      console.log(`[API:${requestId}] SUCCESS - Returning ${responseSizeMB}MB payload`);
-
-      // Create response with explicit headers to handle large payloads
-      const resp = NextResponse.json<GenerateResponse>(responsePayload);
-      resp.headers.set('Content-Type', 'application/json');
-      resp.headers.set('Content-Length', responseSize.toString());
-
-      return resp;
-    }
-  }
-
-  // If no image found, check for text error
-  for (const part of parts) {
-    if (part.text) {
-      console.error(`[API:${requestId}] Gemini returned text instead of image: ${part.text.substring(0, 100)}`);
-      return NextResponse.json<GenerateResponse>(
-        {
-          success: false,
-          error: `Model returned text instead of image: ${part.text.substring(0, 200)}`,
-        },
-        { status: 500 }
-      );
-    }
-  }
-
-  console.error(`[API:${requestId}] No image or text found in Gemini response`);
-  return NextResponse.json<GenerateResponse>(
-    {
-      success: false,
-      error: "No image in response",
-    },
-    { status: 500 }
-  );
-}
 
 
 /**