@ai-sdk/alibaba 1.0.34 → 1.0.35

package/CHANGELOG.md

@@ -1,5 +1,14 @@
 # @ai-sdk/alibaba
 
+## 1.0.35
+
+### Patch Changes
+
+- 7dea716: feat(alibaba): support wan2.7 text-to-video and reference-to-video models with the new protocol (`input.media`, `resolution` + `ratio`)
+- Updated dependencies [0952964]
+  - @ai-sdk/provider-utils@4.0.36
+  - @ai-sdk/openai-compatible@2.0.57
+
 ## 1.0.34
 
 ### Patch Changes

package/dist/index.d.mts

@@ -29,7 +29,7 @@ declare const alibabaEmbeddingModelOptions: z.ZodObject<{
 }, z.core.$strip>;
 type AlibabaEmbeddingModelOptions = z.infer<typeof alibabaEmbeddingModelOptions>;
 
-type AlibabaVideoModelId = 'wan2.6-t2v' | 'wan2.5-t2v-preview' | 'wan2.6-i2v' | 'wan2.6-i2v-flash' | 'wan2.6-r2v' | 'wan2.6-r2v-flash' | (string & {});
+type AlibabaVideoModelId = 'wan2.6-t2v' | 'wan2.5-t2v-preview' | 'wan2.7-t2v' | 'wan2.7-t2v-2026-06-12' | 'wan2.6-i2v' | 'wan2.6-i2v-flash' | 'wan2.6-r2v' | 'wan2.6-r2v-flash' | 'wan2.7-r2v' | 'wan2.7-r2v-2026-06-12' | (string & {});
 
 interface AlibabaProvider extends ProviderV3 {
     (modelId: AlibabaChatModelId): LanguageModelV3;
@@ -135,6 +135,23 @@ type AlibabaVideoModelOptions = {
      * Use character identifiers (character1, character2) in prompts to reference them.
      */
     referenceUrls?: string[] | null;
+    /**
+     * Explicit media array for reference-to-video mode (wan2.7 models).
+     * Overrides the automatic mapping from `inputReferences` and `frameImages`.
+     * Use `Image 1`, `Video 1`, etc. in prompts to reference media items
+     * (images and videos are counted separately, in array order).
+     */
+    media?: Array<{
+        type: 'reference_image' | 'reference_video' | 'first_frame';
+        /** Public URL, or a `data:{mime};base64,{data}` URI for images. */
+        url: string;
+        /** URL to an audio file used as voice reference for this media item. */
+        referenceVoice?: string | null;
+    }> | null;
+    /**
+     * Aspect ratio (wan2.7 text-to-video and reference-to-video models).
+     */
+    ratio?: '16:9' | '9:16' | '1:1' | '4:3' | '3:4' | null;
     /** Polling interval in milliseconds. Defaults to 5000 (5 seconds). */
     pollIntervalMs?: number | null;
     /** Maximum wait time in milliseconds for video generation. Defaults to 600000 (10 minutes). */

package/dist/index.d.ts

@@ -29,7 +29,7 @@ declare const alibabaEmbeddingModelOptions: z.ZodObject<{
 }, z.core.$strip>;
 type AlibabaEmbeddingModelOptions = z.infer<typeof alibabaEmbeddingModelOptions>;
 
-type AlibabaVideoModelId = 'wan2.6-t2v' | 'wan2.5-t2v-preview' | 'wan2.6-i2v' | 'wan2.6-i2v-flash' | 'wan2.6-r2v' | 'wan2.6-r2v-flash' | (string & {});
+type AlibabaVideoModelId = 'wan2.6-t2v' | 'wan2.5-t2v-preview' | 'wan2.7-t2v' | 'wan2.7-t2v-2026-06-12' | 'wan2.6-i2v' | 'wan2.6-i2v-flash' | 'wan2.6-r2v' | 'wan2.6-r2v-flash' | 'wan2.7-r2v' | 'wan2.7-r2v-2026-06-12' | (string & {});
 
 interface AlibabaProvider extends ProviderV3 {
     (modelId: AlibabaChatModelId): LanguageModelV3;
@@ -135,6 +135,23 @@ type AlibabaVideoModelOptions = {
      * Use character identifiers (character1, character2) in prompts to reference them.
      */
     referenceUrls?: string[] | null;
+    /**
+     * Explicit media array for reference-to-video mode (wan2.7 models).
+     * Overrides the automatic mapping from `inputReferences` and `frameImages`.
+     * Use `Image 1`, `Video 1`, etc. in prompts to reference media items
+     * (images and videos are counted separately, in array order).
+     */
+    media?: Array<{
+        type: 'reference_image' | 'reference_video' | 'first_frame';
+        /** Public URL, or a `data:{mime};base64,{data}` URI for images. */
+        url: string;
+        /** URL to an audio file used as voice reference for this media item. */
+        referenceVoice?: string | null;
+    }> | null;
+    /**
+     * Aspect ratio (wan2.7 text-to-video and reference-to-video models).
+     */
+    ratio?: '16:9' | '9:16' | '1:1' | '4:3' | '3:4' | null;
     /** Polling interval in milliseconds. Defaults to 5000 (5 seconds). */
     pollIntervalMs?: number | null;
     /** Maximum wait time in milliseconds for video generation. Defaults to 600000 (10 minutes). */

package/dist/index.js

@@ -860,6 +860,18 @@ var alibabaVideoModelOptionsSchema = (0, import_provider_utils5.lazySchema)(
       watermark: import_v46.z.boolean().nullish(),
       audio: import_v46.z.boolean().nullish(),
       referenceUrls: import_v46.z.array(import_v46.z.string()).nullish(),
+      media: import_v46.z.array(
+        import_v46.z.object({
+          type: import_v46.z.enum([
+            "reference_image",
+            "reference_video",
+            "first_frame"
+          ]),
+          url: import_v46.z.string(),
+          referenceVoice: import_v46.z.string().nullish()
+        })
+      ).nullish(),
+      ratio: import_v46.z.enum(["16:9", "9:16", "1:1", "4:3", "3:4"]).nullish(),
       pollIntervalMs: import_v46.z.number().positive().nullish(),
       pollTimeoutMs: import_v46.z.number().positive().nullish()
     }).passthrough()
@@ -907,6 +919,38 @@ function detectMode(modelId) {
   if (modelId.includes("-r2v")) return "r2v";
   return "t2v";
 }
+function isWan27Model(modelId) {
+  return modelId.startsWith("wan2.7");
+}
+var resolutionTierMap = {
+  "1280x720": "720P",
+  "720x1280": "720P",
+  "960x960": "720P",
+  "1088x832": "720P",
+  "832x1088": "720P",
+  "1920x1080": "1080P",
+  "1080x1920": "1080P",
+  "1440x1440": "1080P",
+  "1632x1248": "1080P",
+  "1248x1632": "1080P",
+  "832x480": "480P",
+  "480x832": "480P",
+  "624x624": "480P"
+};
+var supportedRatios = /* @__PURE__ */ new Set(["16:9", "9:16", "1:1", "4:3", "3:4"]);
+function deriveRatioFromResolution(resolution) {
+  const [width, height] = resolution.split("x").map(Number);
+  if (!Number.isInteger(width) || !Number.isInteger(height) || width <= 0 || height <= 0) {
+    return void 0;
+  }
+  let a = width;
+  let b = height;
+  while (b !== 0) {
+    [a, b] = [b, a % b];
+  }
+  const ratio = `${width / a}:${height / a}`;
+  return supportedRatios.has(ratio) ? ratio : void 0;
+}
 function fileToImageString(file) {
   if (file.type === "url") {
     return file.url;
@@ -921,6 +965,47 @@ function resolveStartImage(options) {
   var _a;
   return (_a = getFirstFrameImage(options)) != null ? _a : options.image;
 }
+function isVideoUrl(url) {
+  return /\.(mp4|mov)([?#]|$)/i.test(url);
+}
+function resolveMedia(options, alibabaOptions, warnings) {
+  var _a;
+  if ((alibabaOptions == null ? void 0 : alibabaOptions.media) != null && alibabaOptions.media.length > 0) {
+    return alibabaOptions.media.map((item) => ({
+      type: item.type,
+      url: item.url,
+      ...item.referenceVoice != null ? { reference_voice: item.referenceVoice } : {}
+    }));
+  }
+  const media = [];
+  for (const reference of (_a = options.inputReferences) != null ? _a : []) {
+    if (reference.type === "url") {
+      media.push({
+        type: isVideoUrl(reference.url) ? "reference_video" : "reference_image",
+        url: reference.url
+      });
+    } else if (reference.mediaType.startsWith("image/")) {
+      media.push({
+        type: "reference_image",
+        url: (0, import_provider_utils5.convertImageModelFileToDataUri)(reference)
+      });
+    } else {
+      warnings.push({
+        type: "unsupported",
+        feature: "inputReferences",
+        details: "Alibaba reference-to-video requires URL references for videos. Non-URL video reference was skipped."
+      });
+    }
+  }
+  const firstFrame = getFirstFrameImage(options);
+  if (firstFrame != null) {
+    media.push({
+      type: "first_frame",
+      url: (0, import_provider_utils5.convertImageModelFileToDataUri)(firstFrame)
+    });
+  }
+  return media.length > 0 ? media : void 0;
+}
 function resolveReferenceUrls(options, alibabaOptions, warnings) {
   var _a;
   if (options.frameImages != null && options.frameImages.length > 0) {
@@ -954,7 +1039,7 @@ var AlibabaVideoModel = class {
     return this.config.provider;
   }
   async doGenerate(options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n;
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p;
     const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
     const warnings = [];
     const mode = detectMode(this.modelId);
@@ -974,16 +1059,27 @@ var AlibabaVideoModel = class {
       input.audio_url = alibabaOptions.audioUrl;
     }
     const startImage = resolveStartImage(options);
-    const referenceUrls = resolveReferenceUrls(
-      options,
-      alibabaOptions,
-      warnings
-    );
+    const wan27 = isWan27Model(this.modelId);
+    const supportsRatio = wan27 && mode !== "i2v";
     if (mode === "i2v" && startImage != null) {
       input.img_url = fileToImageString(startImage);
     }
-    if (mode === "r2v" && referenceUrls != null && referenceUrls.length > 0) {
-      input.reference_urls = referenceUrls;
+    if (mode === "r2v") {
+      if (wan27) {
+        const media = resolveMedia(options, alibabaOptions, warnings);
+        if (media != null) {
+          input.media = media;
+        }
+      } else {
+        const referenceUrls = resolveReferenceUrls(
+          options,
+          alibabaOptions,
+          warnings
+        );
+        if (referenceUrls != null && referenceUrls.length > 0) {
+          input.reference_urls = referenceUrls;
+        }
+      }
     }
     const lastFrame = (_e = (_d = options.frameImages) == null ? void 0 : _d.find(
       (frame) => frame.frameType === "last_frame"
@@ -1010,41 +1106,57 @@ var AlibabaVideoModel = class {
       parameters.seed = options.seed;
     }
     if (options.resolution != null) {
-      if (mode === "i2v") {
-        const resolutionMap = {
-          "1280x720": "720P",
-          "720x1280": "720P",
-          "960x960": "720P",
-          "1088x832": "720P",
-          "832x1088": "720P",
-          "1920x1080": "1080P",
-          "1080x1920": "1080P",
-          "1440x1440": "1080P",
-          "1632x1248": "1080P",
-          "1248x1632": "1080P",
-          "832x480": "480P",
-          "480x832": "480P",
-          "624x624": "480P"
-        };
-        parameters.resolution = resolutionMap[options.resolution] || options.resolution;
+      if (mode === "i2v" || wan27) {
+        const resolutionTier = resolutionTierMap[options.resolution] || options.resolution;
+        if (wan27 && resolutionTier !== "720P" && resolutionTier !== "1080P") {
+          warnings.push({
+            type: "unsupported",
+            feature: "resolution",
+            details: `wan2.7 models only support 720P and 1080P resolutions. The resolution "${options.resolution}" was ignored.`
+          });
+        } else {
+          parameters.resolution = resolutionTier;
+        }
       } else {
         parameters.size = options.resolution.replace("x", "*");
       }
     }
+    if (supportsRatio) {
+      const ratio = (_g = (_f = alibabaOptions == null ? void 0 : alibabaOptions.ratio) != null ? _f : options.aspectRatio) != null ? _g : options.resolution != null ? deriveRatioFromResolution(options.resolution) : void 0;
+      if (ratio != null) {
+        parameters.ratio = ratio;
+      }
+    }
     if ((alibabaOptions == null ? void 0 : alibabaOptions.promptExtend) != null) {
       parameters.prompt_extend = alibabaOptions.promptExtend;
     }
     if ((alibabaOptions == null ? void 0 : alibabaOptions.shotType) != null) {
-      parameters.shot_type = alibabaOptions.shotType;
+      if (wan27) {
+        warnings.push({
+          type: "unsupported",
+          feature: "shotType",
+          details: "wan2.7 models do not support the shotType option. Describe the shot structure in the prompt instead."
+        });
+      } else {
+        parameters.shot_type = alibabaOptions.shotType;
+      }
     }
     if ((alibabaOptions == null ? void 0 : alibabaOptions.watermark) != null) {
       parameters.watermark = alibabaOptions.watermark;
     }
-    const audio = (_f = options.generateAudio) != null ? _f : alibabaOptions == null ? void 0 : alibabaOptions.audio;
+    const audio = (_h = options.generateAudio) != null ? _h : alibabaOptions == null ? void 0 : alibabaOptions.audio;
     if (audio != null) {
-      parameters.audio = audio;
+      if (wan27) {
+        warnings.push({
+          type: "unsupported",
+          feature: "generateAudio",
+          details: "wan2.7 models always generate audio. The audio option was ignored."
+        });
+      } else {
+        parameters.audio = audio;
+      }
     }
-    if (options.aspectRatio) {
+    if (options.aspectRatio && !supportsRatio) {
       warnings.push({
         type: "unsupported",
         feature: "aspectRatio",
@@ -1086,15 +1198,15 @@ var AlibabaVideoModel = class {
       abortSignal: options.abortSignal,
       fetch: this.config.fetch
     });
-    const taskId = (_g = createResponse.output) == null ? void 0 : _g.task_id;
+    const taskId = (_i = createResponse.output) == null ? void 0 : _i.task_id;
     if (!taskId) {
       throw new import_provider4.AISDKError({
         name: "ALIBABA_VIDEO_GENERATION_ERROR",
         message: `No task_id returned from Alibaba API. Response: ${JSON.stringify(createResponse)}`
       });
     }
-    const pollIntervalMs = (_h = alibabaOptions == null ? void 0 : alibabaOptions.pollIntervalMs) != null ? _h : 5e3;
-    const pollTimeoutMs = (_i = alibabaOptions == null ? void 0 : alibabaOptions.pollTimeoutMs) != null ? _i : 6e5;
+    const pollIntervalMs = (_j = alibabaOptions == null ? void 0 : alibabaOptions.pollIntervalMs) != null ? _j : 5e3;
+    const pollTimeoutMs = (_k = alibabaOptions == null ? void 0 : alibabaOptions.pollTimeoutMs) != null ? _k : 6e5;
     const startTime = Date.now();
     let finalResponse;
     let responseHeaders;
@@ -1120,7 +1232,7 @@ var AlibabaVideoModel = class {
         fetch: this.config.fetch
       });
       responseHeaders = pollHeaders;
-      const taskStatus = (_j = statusResponse.output) == null ? void 0 : _j.task_status;
+      const taskStatus = (_l = statusResponse.output) == null ? void 0 : _l.task_status;
       if (taskStatus === "SUCCEEDED") {
         finalResponse = statusResponse;
         break;
@@ -1128,11 +1240,11 @@ var AlibabaVideoModel = class {
       if (taskStatus === "FAILED" || taskStatus === "CANCELED") {
         throw new import_provider4.AISDKError({
           name: "ALIBABA_VIDEO_GENERATION_FAILED",
-          message: `Video generation ${taskStatus.toLowerCase()}. Task ID: ${taskId}. ${(_l = (_k = statusResponse.output) == null ? void 0 : _k.message) != null ? _l : ""}`
+          message: `Video generation ${taskStatus.toLowerCase()}. Task ID: ${taskId}. ${(_n = (_m = statusResponse.output) == null ? void 0 : _m.message) != null ? _n : ""}`
         });
       }
     }
-    const videoUrl = (_m = finalResponse == null ? void 0 : finalResponse.output) == null ? void 0 : _m.video_url;
+    const videoUrl = (_o = finalResponse == null ? void 0 : finalResponse.output) == null ? void 0 : _o.video_url;
     if (!videoUrl) {
       throw new import_provider4.AISDKError({
         name: "ALIBABA_VIDEO_GENERATION_ERROR",
@@ -1157,7 +1269,7 @@ var AlibabaVideoModel = class {
         alibaba: {
           taskId,
           videoUrl,
-          ...((_n = finalResponse == null ? void 0 : finalResponse.output) == null ? void 0 : _n.actual_prompt) ? { actualPrompt: finalResponse.output.actual_prompt } : {},
+          ...((_p = finalResponse == null ? void 0 : finalResponse.output) == null ? void 0 : _p.actual_prompt) ? { actualPrompt: finalResponse.output.actual_prompt } : {},
           ...(finalResponse == null ? void 0 : finalResponse.usage) ? {
             usage: {
               duration: finalResponse.usage.duration,
@@ -1173,7 +1285,7 @@ var AlibabaVideoModel = class {
 };
 
 // src/version.ts
-var VERSION = true ? "1.0.34" : "0.0.0-test";
+var VERSION = true ? "1.0.35" : "0.0.0-test";
 
 // src/alibaba-provider.ts
 function createAlibaba(options = {}) {