@umituz/react-native-ai-fal-provider 1.0.43 → 1.0.45

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@umituz/react-native-ai-fal-provider",
-  "version": "1.0.43",
+  "version": "1.0.45",
   "description": "FAL AI provider for React Native - implements IAIProvider interface for unified AI generation",
   "main": "./src/index.ts",
   "types": "./src/index.ts",

package/src/domain/constants/feature-models.constants.ts

@@ -30,10 +30,12 @@ export const FAL_IMAGE_FEATURE_MODELS: Record<ImageFeatureType, string> = {
 
 /**
  * FAL model IDs for VIDEO processing features
+ * Vidu Q1 Reference-to-Video supports up to 7 reference images
+ * Perfect for multi-person scenarios like kiss/hug with two different people
  */
 export const FAL_VIDEO_FEATURE_MODELS: Record<VideoFeatureType, string> = {
-  "ai-hug": "fal-ai/wan-25-preview/image-to-video",
-  "ai-kiss": "fal-ai/wan-25-preview/image-to-video",
+  "ai-hug": "fal-ai/vidu/q1/reference-to-video",
+  "ai-kiss": "fal-ai/vidu/q1/reference-to-video",
 };
 
 /**

package/src/domain/types/input-builders.types.ts

@@ -38,6 +38,10 @@ export interface VideoFromImageOptions {
   /** @deprecated Use prompt instead */
   motion_prompt?: string;
   duration?: number;
+  /** Vidu Q1: Video aspect ratio - "16:9", "9:16", or "1:1" */
+  aspect_ratio?: "16:9" | "9:16" | "1:1";
+  /** Vidu Q1: Movement intensity - "auto", "small", "medium", or "large" */
+  movement_amplitude?: "auto" | "small" | "medium" | "large";
 }
 
 export interface FaceSwapOptions {

package/src/infrastructure/services/fal-provider.ts

@@ -27,7 +27,6 @@ import {
   buildPhotoRestoreInput,
   buildFaceSwapInput,
   buildRemoveBackgroundInput,
-  buildRemoveObjectInput,
   buildReplaceBackgroundInput,
   buildKontextStyleTransferInput,
   buildVideoFromImageInput,
@@ -240,9 +239,10 @@ export class FalProvider implements IAIProvider {
   buildVideoFeatureInput(feature: VideoFeatureType, data: VideoFeatureInputData): Record<string, unknown> {
     const { sourceImageBase64, targetImageBase64, prompt, options } = data;
 
+    // Vidu Q1 optimized prompts for reference-to-video with multiple people
     const defaultPrompts: Record<VideoFeatureType, string> = {
-      "ai-kiss": "Two people kissing romantically, smooth natural motion, cinematic quality",
-      "ai-hug": "Two people hugging warmly, gentle embrace, smooth natural motion, cinematic quality",
+      "ai-kiss": "A romantic couple kissing tenderly, the two reference people sharing an intimate kiss moment, smooth natural movement, cinematic lighting, high quality video",
+      "ai-hug": "A heartwarming embrace between two people, the reference characters hugging warmly with genuine emotion, gentle natural movement, cinematic quality, touching moment",
     };
 
     const effectivePrompt = prompt || defaultPrompts[feature] || "Generate video with natural motion";
@@ -250,8 +250,8 @@ export class FalProvider implements IAIProvider {
     return buildVideoFromImageInput(sourceImageBase64, {
       prompt: effectivePrompt,
       target_image: targetImageBase64,
-      duration: options?.duration as number,
-      ...options,
+      aspect_ratio: (options?.aspect_ratio as "16:9" | "9:16" | "1:1") || "9:16",
+      movement_amplitude: (options?.movement_amplitude as "auto" | "small" | "medium" | "large") || "medium",
     });
   }
 }

package/src/infrastructure/utils/input-builders.util.ts

@@ -77,24 +77,29 @@ export function buildPhotoRestoreInput(
 }
 
 /**
- * Build AI hug/kiss video input for FAL wan-25-preview
- * Supports dual images for interaction videos
+ * Build reference-to-video input for FAL Vidu Q1
+ * Supports up to 7 reference images for multi-person scenarios like kiss/hug
+ * Uses reference_image_urls array for consistent character appearance
  */
 export function buildVideoFromImageInput(
   base64: string,
   options?: VideoFromImageOptions,
 ): Record<string, unknown> {
-  const params: Record<string, unknown> = {
-    prompt: options?.prompt || options?.motion_prompt,
-    num_frames: options?.duration ? Math.ceil(options.duration * 24) : undefined,
-  };
+  const formatImage = (b64: string) =>
+    b64.startsWith("data:") ? b64 : `data:image/jpeg;base64,${b64}`;
 
-  // If target image is provided, use dual image format
+  // Build reference images array - both source and target for kiss/hug
+  const referenceImages: string[] = [formatImage(base64)];
   if (options?.target_image) {
-    return buildDualImageInput(base64, options.target_image, params);
+    referenceImages.push(formatImage(options.target_image));
   }
 
-  return buildSingleImageInput(base64, params);
+  return {
+    prompt: options?.prompt || options?.motion_prompt || "Generate natural motion video",
+    reference_image_urls: referenceImages,
+    aspect_ratio: options?.aspect_ratio || "9:16",
+    movement_amplitude: options?.movement_amplitude || "auto",
+  };
 }
 
 /**