@ai-sdk/xai 3.0.119 → 3.0.121

package/CHANGELOG.md

@@ -1,5 +1,18 @@
 # @ai-sdk/xai
 
+## 3.0.121
+
+### Patch Changes
+
+- 70a6c3b: feat (provider/xai): support Grok Imagine Video 1.5. Adds the `grok-imagine-video-1.5` model id and native `1080p` for text-to-video and image-to-video (the standard `resolution: '1920x1080'` now maps to `1080p`). Reference-to-video remains capped at `720p`, so a `1080p` request in that mode is downgraded with a warning. Also fixes reference routing: previously any non-empty `inputReferences` array selected reference-to-video, so an array holding only a non-image reference sent `reference_images: []` with no usable references.
+- 70a6c3b: feat (provider/xai): add `referenceVoiceIds` for reference-to-video reference audio. Pass up to 3 xAI preset voice ids (e.g. `['eve']`) and reference them from the prompt with `<AUDIO_0>`–`<AUDIO_2>`; they are sent as `reference_audios: [{ voice_id }]` on `POST /v1/videos/generations`.
+
+## 3.0.120
+
+### Patch Changes
+
+- 3283e3e: feat(xai): support the priority service tier on chat and responses
+
 ## 3.0.119
 
 ### Patch Changes

package/dist/index.d.mts

@@ -14,6 +14,10 @@ declare const xaiLanguageModelChatOptions: z.ZodObject<{
     }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
+    serviceTier: z.ZodOptional<z.ZodEnum<{
+        default: "default";
+        priority: "priority";
+    }>>;
     parallel_function_calling: z.ZodOptional<z.ZodBoolean>;
     searchParameters: z.ZodOptional<z.ZodObject<{
         mode: z.ZodEnum<{
@@ -87,6 +91,10 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
     }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
+    serviceTier: z.ZodOptional<z.ZodEnum<{
+        default: "default";
+        priority: "priority";
+    }>>;
     store: z.ZodOptional<z.ZodBoolean>;
     previousResponseId: z.ZodOptional<z.ZodString>;
     include: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodEnum<{
@@ -112,11 +120,12 @@ declare const xaiImageModelOptions: z.ZodObject<{
 }, z.core.$strip>;
 type XaiImageModelOptions = z.infer<typeof xaiImageModelOptions>;
 
-type XaiVideoModelId = 'grok-imagine-video' | (string & {});
+type XaiVideoModelId = 'grok-imagine-video' | 'grok-imagine-video-1.5' | (string & {});
 
 declare const resolutionSchema: z.ZodEnum<{
     "480p": "480p";
     "720p": "720p";
+    "1080p": "1080p";
 }>;
 type XaiVideoResolution = z.infer<typeof resolutionSchema>;
 interface XaiVideoSharedOptions {
@@ -153,6 +162,10 @@ interface XaiVideoReferenceToVideoOptions extends XaiVideoSharedOptions, XaiVide
     mode: 'reference-to-video';
     /** Reference image URLs (1-7) for R2V generation. */
     referenceImageUrls: string[];
+    /**
+     * Preset voice ids (up to 3) that give the subject a voice.
+     */
+    referenceVoiceIds?: string[];
 }
 interface XaiVideoGenerationOptions extends XaiVideoSharedOptions, XaiVideoUserOptions {
     mode?: undefined;
@@ -174,6 +187,10 @@ interface XaiLegacyReferenceToVideoOptions extends XaiVideoSharedOptions, XaiVid
      */
     mode?: undefined;
     referenceImageUrls: string[];
+    /**
+     * Preset voice ids (up to 3) that give the subject a voice.
+     */
+    referenceVoiceIds?: string[];
 }
 /**
  * Provider options for xAI video generation.
@@ -185,6 +202,9 @@ interface XaiLegacyReferenceToVideoOptions extends XaiVideoSharedOptions, XaiVid
  * - `'reference-to-video'` + `referenceImageUrls` -- R2V generation  (`POST /v1/videos/generations`)
  * - no `mode`                                     -- standard generation from text prompts or image input
  *
+ * Reference images may also come from the top-level `inputReferences` option
+ * instead of `referenceImageUrls`.
+ *
  * Runtime remains backward compatible with legacy auto-detected provider
  * options, but the public TypeScript type is intentionally explicit so editors
  * can suggest valid modes and flag invalid field combinations.

package/dist/index.d.ts

@@ -14,6 +14,10 @@ declare const xaiLanguageModelChatOptions: z.ZodObject<{
     }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
+    serviceTier: z.ZodOptional<z.ZodEnum<{
+        default: "default";
+        priority: "priority";
+    }>>;
     parallel_function_calling: z.ZodOptional<z.ZodBoolean>;
     searchParameters: z.ZodOptional<z.ZodObject<{
         mode: z.ZodEnum<{
@@ -87,6 +91,10 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
     }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
+    serviceTier: z.ZodOptional<z.ZodEnum<{
+        default: "default";
+        priority: "priority";
+    }>>;
     store: z.ZodOptional<z.ZodBoolean>;
     previousResponseId: z.ZodOptional<z.ZodString>;
     include: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodEnum<{
@@ -112,11 +120,12 @@ declare const xaiImageModelOptions: z.ZodObject<{
 }, z.core.$strip>;
 type XaiImageModelOptions = z.infer<typeof xaiImageModelOptions>;
 
-type XaiVideoModelId = 'grok-imagine-video' | (string & {});
+type XaiVideoModelId = 'grok-imagine-video' | 'grok-imagine-video-1.5' | (string & {});
 
 declare const resolutionSchema: z.ZodEnum<{
     "480p": "480p";
     "720p": "720p";
+    "1080p": "1080p";
 }>;
 type XaiVideoResolution = z.infer<typeof resolutionSchema>;
 interface XaiVideoSharedOptions {
@@ -153,6 +162,10 @@ interface XaiVideoReferenceToVideoOptions extends XaiVideoSharedOptions, XaiVide
     mode: 'reference-to-video';
     /** Reference image URLs (1-7) for R2V generation. */
     referenceImageUrls: string[];
+    /**
+     * Preset voice ids (up to 3) that give the subject a voice.
+     */
+    referenceVoiceIds?: string[];
 }
 interface XaiVideoGenerationOptions extends XaiVideoSharedOptions, XaiVideoUserOptions {
     mode?: undefined;
@@ -174,6 +187,10 @@ interface XaiLegacyReferenceToVideoOptions extends XaiVideoSharedOptions, XaiVid
      */
     mode?: undefined;
     referenceImageUrls: string[];
+    /**
+     * Preset voice ids (up to 3) that give the subject a voice.
+     */
+    referenceVoiceIds?: string[];
 }
 /**
  * Provider options for xAI video generation.
@@ -185,6 +202,9 @@ interface XaiLegacyReferenceToVideoOptions extends XaiVideoSharedOptions, XaiVid
  * - `'reference-to-video'` + `referenceImageUrls` -- R2V generation  (`POST /v1/videos/generations`)
  * - no `mode`                                     -- standard generation from text prompts or image input
  *
+ * Reference images may also come from the top-level `inputReferences` option
+ * instead of `referenceImageUrls`.
+ *
  * Runtime remains backward compatible with legacy auto-detected provider
  * options, but the public TypeScript type is intentionally explicit so editors
  * can suggest valid modes and flag invalid field combinations.

package/dist/index.js

@@ -277,6 +277,7 @@ var xaiLanguageModelChatOptions = import_v42.z.object({
   reasoningEffort: import_v42.z.enum(["none", "low", "medium", "high", "xhigh"]).optional(),
   logprobs: import_v42.z.boolean().optional(),
   topLogprobs: import_v42.z.number().int().min(0).max(8).optional(),
+  serviceTier: import_v42.z.enum(["default", "priority"]).optional(),
   /**
    * Whether to enable parallel function calling during tool use.
    * When true, the model can call multiple functions in parallel.
@@ -501,6 +502,8 @@ var XaiChatLanguageModel = class {
       top_p: topP,
       seed,
       reasoning_effort: options.reasoningEffort,
+      // scheduling priority
+      service_tier: options.serviceTier,
       // parallel function calling
       parallel_function_calling: options.parallel_function_calling,
       // response format
@@ -635,6 +638,11 @@ var XaiChatLanguageModel = class {
         inputTokens: { total: 0, noCache: 0, cacheRead: 0, cacheWrite: 0 },
         outputTokens: { total: 0, text: 0, reasoning: 0 }
       },
+      ...response.service_tier != null && {
+        providerMetadata: {
+          xai: { serviceTier: response.service_tier }
+        }
+      },
       request: { body },
       response: {
         ...getResponseMetadata(response),
@@ -703,6 +711,7 @@ var XaiChatLanguageModel = class {
       raw: void 0
     };
     let usage = void 0;
+    let serviceTier = void 0;
     let isFirstChunk = true;
     const contentBlocks = {};
     const lastReasoningDeltas = {};
@@ -743,6 +752,9 @@ var XaiChatLanguageModel = class {
             if (value.usage != null) {
               usage = convertXaiChatUsage(value.usage);
             }
+            if (value.service_tier != null) {
+              serviceTier = value.service_tier;
+            }
             const choice = value.choices[0];
             if ((choice == null ? void 0 : choice.finish_reason) != null) {
               finishReason = {
@@ -857,6 +869,9 @@ var XaiChatLanguageModel = class {
                   cacheWrite: 0
                 },
                 outputTokens: { total: 0, text: 0, reasoning: 0 }
+              },
+              ...serviceTier != null && {
+                providerMetadata: { xai: { serviceTier } }
               }
             });
           }
@@ -912,6 +927,7 @@ var xaiChatResponseSchema = import_v44.z.object({
   object: import_v44.z.literal("chat.completion").nullish(),
   usage: xaiUsageSchema.nullish(),
   citations: import_v44.z.array(import_v44.z.string().url()).nullish(),
+  service_tier: import_v44.z.string().nullish(),
   code: import_v44.z.string().nullish(),
   error: import_v44.z.string().nullish()
 });
@@ -941,7 +957,8 @@ var xaiChatChunkSchema = import_v44.z.object({
     })
   ),
   usage: xaiUsageSchema.nullish(),
-  citations: import_v44.z.array(import_v44.z.string().url()).nullish()
+  citations: import_v44.z.array(import_v44.z.string().url()).nullish(),
+  service_tier: import_v44.z.string().nullish()
 });
 var xaiStreamErrorSchema = import_v44.z.object({
   code: import_v44.z.string(),
@@ -1509,7 +1526,8 @@ var xaiResponsesResponseSchema = import_v47.z.object({
   object: import_v47.z.literal("response"),
   output: import_v47.z.array(outputItemSchema),
   usage: xaiResponsesUsageSchema.nullish(),
-  status: import_v47.z.string()
+  status: import_v47.z.string(),
+  service_tier: import_v47.z.string().nullish()
 });
 var xaiResponsesChunkSchema = import_v47.z.union([
   import_v47.z.object({
@@ -1777,7 +1795,8 @@ var xaiResponsesChunkSchema = import_v47.z.union([
     type: import_v47.z.literal("response.incomplete"),
     response: import_v47.z.object({
       incomplete_details: import_v47.z.object({ reason: import_v47.z.string() }).nullish(),
-      usage: xaiResponsesUsageSchema.nullish()
+      usage: xaiResponsesUsageSchema.nullish(),
+      service_tier: import_v47.z.string().nullish()
     })
   }),
   import_v47.z.object({
@@ -1820,6 +1839,7 @@ var xaiLanguageModelResponsesOptions = import_v48.z.object({
   reasoningEffort: import_v48.z.enum(["none", "low", "medium", "high", "xhigh"]).optional(),
   logprobs: import_v48.z.boolean().optional(),
   topLogprobs: import_v48.z.number().int().min(0).max(8).optional(),
+  serviceTier: import_v48.z.enum(["default", "priority"]).optional(),
   /**
    * Whether to store the input message(s) and model response for later retrieval.
    * Must be set to `false` for teams with Zero Data Retention (ZDR) enabled,
@@ -2248,6 +2268,9 @@ var XaiResponsesLanguageModel = class {
       },
       ...options.previousResponseId != null && {
         previous_response_id: options.previousResponseId
+      },
+      ...options.serviceTier != null && {
+        service_tier: options.serviceTier
       }
     };
     if (xaiTools2 && xaiTools2.length > 0) {
@@ -2424,6 +2447,11 @@ var XaiResponsesLanguageModel = class {
         inputTokens: { total: 0, noCache: 0, cacheRead: 0, cacheWrite: 0 },
         outputTokens: { total: 0, text: 0, reasoning: 0 }
       },
+      ...response.service_tier != null && {
+        providerMetadata: {
+          xai: { serviceTier: response.service_tier }
+        }
+      },
       request: { body },
       response: {
         ...getResponseMetadata(response),
@@ -2465,6 +2493,7 @@ var XaiResponsesLanguageModel = class {
     };
     let hasFunctionCall = false;
     let usage = void 0;
+    let serviceTier = void 0;
     let isFirstChunk = true;
     const contentBlocks = {};
     const seenToolCalls = /* @__PURE__ */ new Set();
@@ -2478,7 +2507,7 @@ var XaiResponsesLanguageModel = class {
             controller.enqueue({ type: "stream-start", warnings });
           },
           transform(chunk, controller) {
-            var _a2, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o;
+            var _a2, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p;
             if (options.includeRawChunks) {
               controller.enqueue({ type: "raw", rawValue: chunk.rawValue });
             }
@@ -2608,8 +2637,9 @@ var XaiResponsesLanguageModel = class {
               if (response2.usage) {
                 usage = convertXaiResponsesUsage(response2.usage);
               }
+              serviceTier = (_c = response2.service_tier) != null ? _c : void 0;
               if (event.type === "response.incomplete") {
-                const reason = "incomplete_details" in response2 ? (_c = response2.incomplete_details) == null ? void 0 : _c.reason : void 0;
+                const reason = "incomplete_details" in response2 ? (_d = response2.incomplete_details) == null ? void 0 : _d.reason : void 0;
                 finishReason = {
                   unified: reason ? mapXaiResponsesFinishReason(reason) : "other",
                   raw: reason != null ? reason : "incomplete"
@@ -2623,7 +2653,7 @@ var XaiResponsesLanguageModel = class {
               return;
             }
             if (event.type === "response.failed") {
-              const reason = (_d = event.response.incomplete_details) == null ? void 0 : _d.reason;
+              const reason = (_e = event.response.incomplete_details) == null ? void 0 : _e.reason;
               finishReason = {
                 unified: reason ? mapXaiResponsesFinishReason(reason) : "error",
                 raw: reason != null ? reason : "error"
@@ -2719,13 +2749,13 @@ var XaiResponsesLanguageModel = class {
                     toolCallId: part.id,
                     toolName,
                     result: {
-                      queries: (_e = part.queries) != null ? _e : [],
-                      results: (_g = (_f = part.results) == null ? void 0 : _f.map((result) => ({
+                      queries: (_f = part.queries) != null ? _f : [],
+                      results: (_h = (_g = part.results) == null ? void 0 : _g.map((result) => ({
                         fileId: result.file_id,
                         filename: result.filename,
                         score: result.score,
                         text: result.text
-                      }))) != null ? _g : null
+                      }))) != null ? _h : null
                     }
                   });
                 }
@@ -2743,17 +2773,17 @@ var XaiResponsesLanguageModel = class {
                   "x_semantic_search",
                   "x_thread_fetch"
                 ];
-                let toolName = (_h = part.name) != null ? _h : "";
-                if (webSearchSubTools.includes((_i = part.name) != null ? _i : "") || part.type === "web_search_call") {
+                let toolName = (_i = part.name) != null ? _i : "";
+                if (webSearchSubTools.includes((_j = part.name) != null ? _j : "") || part.type === "web_search_call") {
                   toolName = webSearchToolName != null ? webSearchToolName : "web_search";
-                } else if (xSearchSubTools.includes((_j = part.name) != null ? _j : "") || part.type === "x_search_call") {
+                } else if (xSearchSubTools.includes((_k = part.name) != null ? _k : "") || part.type === "x_search_call") {
                   toolName = xSearchToolName != null ? xSearchToolName : "x_search";
                 } else if (part.name === "code_execution" || part.type === "code_interpreter_call" || part.type === "code_execution_call") {
                   toolName = codeExecutionToolName != null ? codeExecutionToolName : "code_execution";
                 } else if (part.type === "mcp_call") {
-                  toolName = (_k = mcpToolName != null ? mcpToolName : part.name) != null ? _k : "mcp";
+                  toolName = (_l = mcpToolName != null ? mcpToolName : part.name) != null ? _l : "mcp";
                 }
-                const toolInput = part.type === "custom_tool_call" ? (_l = part.input) != null ? _l : "" : part.type === "mcp_call" ? (_m = part.arguments) != null ? _m : "" : (_n = part.arguments) != null ? _n : "";
+                const toolInput = part.type === "custom_tool_call" ? (_m = part.input) != null ? _m : "" : part.type === "mcp_call" ? (_n = part.arguments) != null ? _n : "" : (_o = part.arguments) != null ? _o : "";
                 const shouldEmit = part.type === "custom_tool_call" ? event.type === "response.output_item.done" : !seenToolCalls.has(part.id);
                 if (shouldEmit && !seenToolCalls.has(part.id)) {
                   seenToolCalls.add(part.id);
@@ -2814,7 +2844,7 @@ var XaiResponsesLanguageModel = class {
                           sourceType: "url",
                           id: self.config.generateId(),
                           url: annotation.url,
-                          title: (_o = annotation.title) != null ? _o : annotation.url
+                          title: (_p = annotation.title) != null ? _p : annotation.url
                         });
                       }
                     }
@@ -2868,6 +2898,9 @@ var XaiResponsesLanguageModel = class {
                   cacheWrite: 0
                 },
                 outputTokens: { total: 0, text: 0, reasoning: 0 }
+              },
+              ...serviceTier != null && {
+                providerMetadata: { xai: { serviceTier } }
               }
             });
           }
@@ -2934,7 +2967,7 @@ var xaiTools = {
 };
 
 // src/version.ts
-var VERSION = true ? "3.0.119" : "0.0.0-test";
+var VERSION = true ? "3.0.121" : "0.0.0-test";
 
 // src/xai-video-model.ts
 var import_provider6 = require("@ai-sdk/provider");
@@ -2945,7 +2978,7 @@ var import_v417 = require("zod/v4");
 var import_provider_utils15 = require("@ai-sdk/provider-utils");
 var import_v416 = require("zod/v4");
 var nonEmptyStringSchema = import_v416.z.string().min(1);
-var resolutionSchema = import_v416.z.enum(["480p", "720p"]);
+var resolutionSchema = import_v416.z.enum(["480p", "720p", "1080p"]);
 var modeSchema = import_v416.z.enum(["edit-video", "extend-video", "reference-to-video"]);
 var baseFields = {
   pollIntervalMs: import_v416.z.number().positive().nullish(),
@@ -2955,22 +2988,28 @@ var baseFields = {
 var userField = {
   user: import_v416.z.string().optional()
 };
+var referenceVoiceIdsField = {
+  referenceVoiceIds: import_v416.z.array(nonEmptyStringSchema).max(3).optional()
+};
 var editVideoSchema = import_v416.z.object({
   ...baseFields,
   ...userField,
   mode: import_v416.z.literal("edit-video"),
   videoUrl: nonEmptyStringSchema,
-  referenceImageUrls: import_v416.z.undefined().optional()
+  referenceImageUrls: import_v416.z.undefined().optional(),
+  referenceVoiceIds: import_v416.z.undefined().optional()
 });
 var extendVideoSchema = import_v416.z.object({
   ...baseFields,
   mode: import_v416.z.literal("extend-video"),
   videoUrl: nonEmptyStringSchema,
-  referenceImageUrls: import_v416.z.undefined().optional()
+  referenceImageUrls: import_v416.z.undefined().optional(),
+  referenceVoiceIds: import_v416.z.undefined().optional()
 });
 var referenceToVideoSchema = import_v416.z.object({
   ...baseFields,
   ...userField,
+  ...referenceVoiceIdsField,
   mode: import_v416.z.literal("reference-to-video"),
   referenceImageUrls: import_v416.z.array(nonEmptyStringSchema).min(1).max(7),
   videoUrl: import_v416.z.undefined().optional()
@@ -2978,6 +3017,7 @@ var referenceToVideoSchema = import_v416.z.object({
 var autoDetectSchema = import_v416.z.object({
   ...baseFields,
   ...userField,
+  ...referenceVoiceIdsField,
   mode: import_v416.z.undefined().optional(),
   videoUrl: nonEmptyStringSchema.optional(),
   referenceImageUrls: import_v416.z.array(nonEmptyStringSchema).min(1).max(7).optional()
@@ -2992,6 +3032,7 @@ var runtimeSchema = import_v416.z.object({
   mode: modeSchema.optional(),
   videoUrl: nonEmptyStringSchema.optional(),
   referenceImageUrls: import_v416.z.array(nonEmptyStringSchema).min(1).max(7).optional(),
+  referenceVoiceIds: import_v416.z.array(nonEmptyStringSchema).max(3).optional(),
   user: import_v416.z.string().optional(),
   ...baseFields
 }).passthrough();
@@ -3001,6 +3042,7 @@ var xaiVideoModelOptionsSchema = (0, import_provider_utils15.lazySchema)(
 
 // src/xai-video-model.ts
 var RESOLUTION_MAP = {
+  "1920x1080": "1080p",
   "1280x720": "720p",
   "854x480": "480p",
   "640x480": "480p"
@@ -3022,6 +3064,7 @@ function getTopLevelMediaType(mediaType) {
   return slashIndex === -1 ? mediaType : mediaType.substring(0, slashIndex);
 }
 var isVideoFile = (file) => file.mediaType != null && getTopLevelMediaType(file.mediaType) === "video";
+var isImageReference = (file) => file.mediaType == null || getTopLevelMediaType(file.mediaType) === "image";
 function fileToXaiImageUrl(file) {
   if (file.type === "url") {
     return file.url;
@@ -3031,26 +3074,29 @@ function fileToXaiImageUrl(file) {
 }
 function resolveReferenceImages(options, xaiOptions, warnings) {
   if (options.inputReferences != null && options.inputReferences.length > 0) {
-    const imageReferences = options.inputReferences.filter((reference) => {
-      if (isVideoFile(reference)) {
+    const imageFiles = [];
+    for (const reference of options.inputReferences) {
+      if (!isImageReference(reference)) {
         warnings.push({
           type: "unsupported",
           feature: "inputReferences",
-          details: 'xAI reference-to-video accepts image references only. The video reference was ignored. Use providerOptions.xai.mode "extend-video" to continue from a video.'
+          details: isVideoFile(reference) ? 'xAI reference-to-video accepts image references only. The video reference was ignored. Use providerOptions.xai.mode "extend-video" to continue from a video.' : "xAI reference-to-video accepts image references only. The non-image reference was ignored."
         });
-        return false;
+        continue;
       }
-      return true;
-    });
-    return imageReferences.map((reference) => ({
-      url: fileToXaiImageUrl(reference)
-    }));
+      imageFiles.push(reference);
+    }
+    return imageFiles.length > 0 ? imageFiles.map((reference) => ({ url: fileToXaiImageUrl(reference) })) : void 0;
   }
   if ((xaiOptions == null ? void 0 : xaiOptions.referenceImageUrls) != null && xaiOptions.referenceImageUrls.length > 0) {
     return xaiOptions.referenceImageUrls.map((url) => ({ url }));
   }
   return void 0;
 }
+function hasImageInputReference(options) {
+  var _a, _b;
+  return (_b = (_a = options.inputReferences) == null ? void 0 : _a.some(isImageReference)) != null ? _b : false;
+}
 function resolveVideoMode(options, xaiOptions) {
   if ((xaiOptions == null ? void 0 : xaiOptions.mode) != null) {
     return xaiOptions.mode;
@@ -3059,9 +3105,8 @@ function resolveVideoMode(options, xaiOptions) {
     return "edit-video";
   }
   const hasFrameImages = options.frameImages != null && options.frameImages.length > 0;
-  const hasInputReferences = options.inputReferences != null && options.inputReferences.length > 0;
   const hasLegacyReferenceUrls = (xaiOptions == null ? void 0 : xaiOptions.referenceImageUrls) != null && xaiOptions.referenceImageUrls.length > 0;
-  if (!hasFrameImages && (hasInputReferences || hasLegacyReferenceUrls)) {
+  if (!hasFrameImages && (hasImageInputReference(options) || hasLegacyReferenceUrls)) {
     return "reference-to-video";
   }
   return void 0;
@@ -3168,7 +3213,7 @@ var XaiVideoModel = class {
         warnings.push({
           type: "unsupported",
           feature: "resolution",
-          details: `Unrecognized resolution "${options.resolution}". Use providerOptions.xai.resolution with "480p" or "720p" instead.`
+          details: `Unrecognized resolution "${options.resolution}". Use providerOptions.xai.resolution with "480p", "720p", or "1080p" instead.`
         });
       }
     }
@@ -3207,13 +3252,47 @@ var XaiVideoModel = class {
       );
       if (referenceImages != null) {
         body.reference_images = referenceImages;
+      } else {
+        warnings.push({
+          type: "unsupported",
+          feature: "referenceImages",
+          details: "xAI reference-to-video requires at least one image reference. The video will be generated without reference images."
+        });
+      }
+      const referenceVoiceIds = xaiOptions == null ? void 0 : xaiOptions.referenceVoiceIds;
+      if (referenceVoiceIds != null && referenceVoiceIds.length > 0) {
+        body.reference_audios = referenceVoiceIds.map((voiceId) => ({
+          voice_id: voiceId
+        }));
+      }
+      if (body.resolution === "1080p") {
+        warnings.push({
+          type: "unsupported",
+          feature: "resolution",
+          details: "xAI reference-to-video is limited to 720p. The request was downgraded from 1080p to 720p."
+        });
+        body.resolution = "720p";
       }
     }
+    if (body.resolution === "1080p" && this.modelId === "grok-imagine-video") {
+      warnings.push({
+        type: "unsupported",
+        feature: "resolution",
+        details: 'xAI model "grok-imagine-video" does not support 1080p. Use "grok-imagine-video-1.5" for 1080p, or a lower resolution. The request was sent with 1080p.'
+      });
+    }
     if (options.inputReferences != null && options.inputReferences.length > 0 && !hasReferenceImages) {
       warnings.push({
         type: "unsupported",
         feature: "inputReferences",
-        details: "xAI only supports inputReferences for reference-to-video generation. The reference images were ignored."
+        details: hasImageInputReference(options) ? "xAI only supports inputReferences for reference-to-video generation. The reference images were ignored." : "xAI reference-to-video requires at least one image reference. The references were ignored."
+      });
+    }
+    if ((xaiOptions == null ? void 0 : xaiOptions.referenceVoiceIds) != null && xaiOptions.referenceVoiceIds.length > 0 && !hasReferenceImages) {
+      warnings.push({
+        type: "unsupported",
+        feature: "referenceVoiceIds",
+        details: "xAI only supports reference voices for reference-to-video generation. The reference voices were ignored."
       });
     }
     if (!isExtension && (xaiOptions == null ? void 0 : xaiOptions.user) !== void 0) {
@@ -3228,6 +3307,7 @@ var XaiVideoModel = class {
           "resolution",
           "videoUrl",
           "referenceImageUrls",
+          "referenceVoiceIds",
           "user"
         ].includes(key)) {
           body[key] = value;