ai 6.0.221 → 6.0.222

package/CHANGELOG.md

@@ -1,5 +1,18 @@
 # ai
 
+## 6.0.222
+
+### Patch Changes
+
+- e1af05f: feat (video): support video (not just image) reference inputs in `inputReferences` for reference-to-video generation
+- Updated dependencies [2bfb16a]
+- Updated dependencies [34b5acc]
+- Updated dependencies [e1af05f]
+- Updated dependencies [1ce0d1c]
+  - @ai-sdk/gateway@3.0.146
+  - @ai-sdk/provider@3.0.14
+  - @ai-sdk/provider-utils@4.0.38
+
 ## 6.0.221
 
 ### Patch Changes

package/dist/index.d.mts

@@ -5930,9 +5930,22 @@ declare function experimental_generateVideo({ model: modelArg, prompt: promptArg
         frameType: Experimental_VideoModelV3FrameType;
     }>;
     /**
-     * Reference image inputs for reference-to-video generation.
+     * Reference inputs for reference-to-video generation.
+     *
+     * Each entry may be a plain image/video ({@link DataContent}), or an object
+     * form that carries an explicit `mediaType`.
      */
-    inputReferences?: Array<DataContent>;
+    inputReferences?: Array<DataContent | {
+        /**
+         * The reference image or video.
+         */
+        data: DataContent;
+        /**
+         * The media type of the reference (e.g. 'image/png',
+         * 'video/mp4').
+         */
+        mediaType?: string;
+    }>;
     /**
      * Whether the model should generate audio alongside the video.
      */

package/dist/index.d.ts

@@ -5930,9 +5930,22 @@ declare function experimental_generateVideo({ model: modelArg, prompt: promptArg
         frameType: Experimental_VideoModelV3FrameType;
     }>;
     /**
-     * Reference image inputs for reference-to-video generation.
+     * Reference inputs for reference-to-video generation.
+     *
+     * Each entry may be a plain image/video ({@link DataContent}), or an object
+     * form that carries an explicit `mediaType`.
      */
-    inputReferences?: Array<DataContent>;
+    inputReferences?: Array<DataContent | {
+        /**
+         * The reference image or video.
+         */
+        data: DataContent;
+        /**
+         * The media type of the reference (e.g. 'image/png',
+         * 'video/mp4').
+         */
+        mediaType?: string;
+    }>;
     /**
      * Whether the model should generate audio alongside the video.
      */

package/dist/index.js

@@ -1282,7 +1282,7 @@ function detectMediaType({
 var import_provider_utils3 = require("@ai-sdk/provider-utils");
 
 // src/version.ts
-var VERSION = true ? "6.0.221" : "0.0.0-test";
+var VERSION = true ? "6.0.222" : "0.0.0-test";
 
 // src/util/download/download.ts
 var download = async ({
@@ -11927,13 +11927,14 @@ async function experimental_generateVideo({
     abortSignal
   });
   const { prompt, image } = normalizePrompt2(promptArg);
-  const normalizedFrameImages = frameImages == null ? void 0 : frameImages.map((frame) => ({
-    image: normalizeImageData(frame.image),
-    frameType: frame.frameType
-  }));
-  const normalizedInputReferences = inputReferences == null ? void 0 : inputReferences.map(
-    (reference) => normalizeImageData(reference)
-  );
+  const normalizedFrameImages = frameImages == null ? void 0 : frameImages.flatMap((frame) => {
+    const normalizedImage = normalizeImageData(frame.image);
+    return normalizedImage != null ? [{ image: normalizedImage, frameType: frame.frameType }] : [];
+  });
+  const normalizedInputReferences = inputReferences == null ? void 0 : inputReferences.flatMap((reference) => {
+    const normalized = normalizeReferenceData(reference);
+    return normalized != null ? [normalized] : [];
+  });
   const effectiveInputReferences = normalizedFrameImages != null && normalizedFrameImages.length > 0 ? void 0 : normalizedInputReferences;
   const warnings = [];
   if (normalizedFrameImages != null && normalizedFrameImages.length > 0 && normalizedInputReferences != null && normalizedInputReferences.length > 0) {
@@ -12087,8 +12088,12 @@ function normalizePrompt2(promptArg) {
     image: promptArg.image != null ? normalizeImageData(promptArg.image) : void 0
   };
 }
-function normalizeImageData(dataContent) {
-  var _a22, _b;
+function detectFileMediaType(data, restrictToImages) {
+  var _a22;
+  const detected = restrictToImages ? detectMediaType({ data, signatures: imageMediaTypeSignatures }) : (_a22 = detectMediaType({ data, signatures: imageMediaTypeSignatures })) != null ? _a22 : detectMediaType({ data, signatures: videoMediaTypeSignatures });
+  return detected != null ? detected : "image/png";
+}
+function normalizeImageData(dataContent, { restrictToImages = true } = {}) {
   if (typeof dataContent === "string") {
     if (dataContent.startsWith("http://") || dataContent.startsWith("https://")) {
       return {
@@ -12098,27 +12103,46 @@ function normalizeImageData(dataContent) {
     }
     if (dataContent.startsWith("data:")) {
       const { mediaType, base64Content } = splitDataUrl(dataContent);
+      const data = (0, import_provider_utils35.convertBase64ToUint8Array)(base64Content != null ? base64Content : "");
       return {
         type: "file",
-        mediaType: mediaType != null ? mediaType : "image/png",
-        data: (0, import_provider_utils35.convertBase64ToUint8Array)(base64Content != null ? base64Content : "")
+        mediaType: mediaType != null ? mediaType : detectFileMediaType(data, restrictToImages),
+        data
       };
     }
-    const bytes2 = (0, import_provider_utils35.convertBase64ToUint8Array)(dataContent);
+    const bytes = (0, import_provider_utils35.convertBase64ToUint8Array)(dataContent);
     return {
       type: "file",
-      mediaType: (_a22 = detectMediaType({
-        data: bytes2,
-        signatures: imageMediaTypeSignatures
-      })) != null ? _a22 : "image/png",
-      data: bytes2
+      mediaType: detectFileMediaType(bytes, restrictToImages),
+      data: bytes
     };
   }
-  const bytes = convertDataContentToUint8Array(dataContent);
+  if (dataContent instanceof Uint8Array || dataContent instanceof ArrayBuffer) {
+    const bytes = dataContent instanceof Uint8Array ? dataContent : new Uint8Array(dataContent);
+    return {
+      type: "file",
+      mediaType: detectFileMediaType(bytes, restrictToImages),
+      data: bytes
+    };
+  }
+  return void 0;
+}
+function normalizeReferenceData(reference) {
+  const isObjectForm = typeof reference === "object" && reference != null && !(reference instanceof Uint8Array) && !(reference instanceof ArrayBuffer) && "data" in reference;
+  if (!isObjectForm) {
+    return normalizeImageData(reference, {
+      restrictToImages: false
+    });
+  }
+  const normalized = normalizeImageData(reference.data, {
+    restrictToImages: false
+  });
+  if (normalized == null) {
+    return normalized;
+  }
   return {
-    type: "file",
-    mediaType: (_b = detectMediaType({ data: bytes, signatures: imageMediaTypeSignatures })) != null ? _b : "image/png",
-    data: bytes
+    ...normalized,
+    ...reference.mediaType != null ? { mediaType: reference.mediaType } : {}
   };
 }
 async function invokeModelMaxVideosPerCall(model) {