@ai-sdk/fal 3.0.0-beta.8 → 3.0.0-canary.32

package/src/fal-speech-model.ts

@@ -1,4 +1,4 @@
-import { SpeechModelV4, SharedV4Warning } from '@ai-sdk/provider';
+import type { SpeechModelV4, SharedV4Warning } from '@ai-sdk/provider';
 import {
   combineHeaders,
   createBinaryResponseHandler,
@@ -7,31 +7,15 @@ import {
   getFromApi,
   parseProviderOptions,
   postJsonToApi,
+  serializeModelOptions,
+  WORKFLOW_SERIALIZE,
+  WORKFLOW_DESERIALIZE,
 } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
-import { FalConfig } from './fal-config';
+import type { FalConfig } from './fal-config';
 import { falFailedResponseHandler } from './fal-error';
-import { FAL_EMOTIONS, FAL_LANGUAGE_BOOSTS } from './fal-api-types';
-import { FalSpeechModelId } from './fal-speech-settings';
-
-const falSpeechModelOptionsSchema = z.looseObject({
-  voice_setting: z
-    .object({
-      speed: z.number().nullish(),
-      vol: z.number().nullish(),
-      voice_id: z.string().nullish(),
-      pitch: z.number().nullish(),
-      english_normalization: z.boolean().nullish(),
-      emotion: z.enum(FAL_EMOTIONS).nullish(),
-    })
-    .partial()
-    .nullish(),
-  audio_setting: z.record(z.string(), z.unknown()).nullish(),
-  language_boost: z.enum(FAL_LANGUAGE_BOOSTS).nullish(),
-  pronunciation_dict: z.record(z.string(), z.string()).nullish(),
-});
-
-export type FalSpeechModelOptions = z.infer<typeof falSpeechModelOptionsSchema>;
+import { falSpeechModelOptionsSchema } from './fal-speech-model-options';
+import type { FalSpeechModelId } from './fal-speech-settings';
 
 interface FalSpeechModelConfig extends FalConfig {
   _internal?: {
@@ -46,6 +30,20 @@ export class FalSpeechModel implements SpeechModelV4 {
     return this.config.provider;
   }
 
+  static [WORKFLOW_SERIALIZE](model: FalSpeechModel) {
+    return serializeModelOptions({
+      modelId: model.modelId,
+      config: model.config,
+    });
+  }
+
+  static [WORKFLOW_DESERIALIZE](options: {
+    modelId: FalSpeechModelId;
+    config: FalSpeechModelConfig;
+  }) {
+    return new FalSpeechModel(options.modelId, options.config);
+  }
+
   constructor(
     readonly modelId: FalSpeechModelId,
     private readonly config: FalSpeechModelConfig,
@@ -112,7 +110,7 @@ export class FalSpeechModel implements SpeechModelV4 {
         path: `https://fal.run/${this.modelId}`,
         modelId: this.modelId,
       }),
-      headers: combineHeaders(this.config.headers(), options.headers),
+      headers: combineHeaders(this.config.headers?.(), options.headers),
       body: requestBody,
       failedResponseHandler: falFailedResponseHandler,
       successfulResponseHandler: createJsonResponseHandler(

package/src/fal-transcription-model-options.ts

@@ -0,0 +1,43 @@
+import { z } from 'zod/v4';
+
+// https://fal.ai/models/fal-ai/whisper/api?platform=http
+export const falTranscriptionModelOptionsSchema = z.object({
+  /**
+   * Language of the audio file. If set to null, the language will be automatically detected. Defaults to null.
+   *
+   * If translate is selected as the task, the audio will be translated to English, regardless of the language selected.
+   */
+  language: z
+    .union([z.enum(['en']), z.string()])
+    .nullish()
+    .default('en'),
+
+  /**
+   * Whether to diarize the audio file. Defaults to true.
+   */
+  diarize: z.boolean().nullish().default(true),
+
+  /**
+   * Level of the chunks to return. Either segment or word. Default value: "segment"
+   */
+  chunkLevel: z.enum(['segment', 'word']).nullish().default('segment'),
+
+  /**
+   * Version of the model to use. All of the models are the Whisper large variant. Default value: "3"
+   */
+  version: z.enum(['3']).nullish().default('3'),
+
+  /**
+   * Default value: 64
+   */
+  batchSize: z.number().nullish().default(64),
+
+  /**
+   * Number of speakers in the audio file. Defaults to null. If not provided, the number of speakers will be automatically detected.
+   */
+  numSpeakers: z.number().nullable().nullish(),
+});
+
+export type FalTranscriptionModelOptions = z.infer<
+  typeof falTranscriptionModelOptionsSchema
+>;

package/src/fal-transcription-model.ts

@@ -1,7 +1,7 @@
 import {
   AISDKError,
-  TranscriptionModelV4,
-  SharedV4Warning,
+  type TranscriptionModelV4,
+  type SharedV4Warning,
 } from '@ai-sdk/provider';
 import {
   combineHeaders,
@@ -12,54 +12,16 @@ import {
   getFromApi,
   parseProviderOptions,
   postJsonToApi,
+  serializeModelOptions,
+  WORKFLOW_SERIALIZE,
+  WORKFLOW_DESERIALIZE,
 } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
-import { FalConfig } from './fal-config';
+import type { FalConfig } from './fal-config';
 import { falErrorDataSchema, falFailedResponseHandler } from './fal-error';
-import { FalTranscriptionModelId } from './fal-transcription-options';
-import { FalTranscriptionAPITypes } from './fal-api-types';
-
-// https://fal.ai/models/fal-ai/whisper/api?platform=http
-const falTranscriptionModelOptionsSchema = z.object({
-  /**
-   * Language of the audio file. If set to null, the language will be automatically detected. Defaults to null.
-   *
-   * If translate is selected as the task, the audio will be translated to English, regardless of the language selected.
-   */
-  language: z
-    .union([z.enum(['en']), z.string()])
-    .nullish()
-    .default('en'),
-
-  /**
-   * Whether to diarize the audio file. Defaults to true.
-   */
-  diarize: z.boolean().nullish().default(true),
-
-  /**
-   * Level of the chunks to return. Either segment or word. Default value: "segment"
-   */
-  chunkLevel: z.enum(['segment', 'word']).nullish().default('segment'),
-
-  /**
-   * Version of the model to use. All of the models are the Whisper large variant. Default value: "3"
-   */
-  version: z.enum(['3']).nullish().default('3'),
-
-  /**
-   * Default value: 64
-   */
-  batchSize: z.number().nullish().default(64),
-
-  /**
-   * Number of speakers in the audio file. Defaults to null. If not provided, the number of speakers will be automatically detected.
-   */
-  numSpeakers: z.number().nullable().nullish(),
-});
-
-export type FalTranscriptionModelOptions = z.infer<
-  typeof falTranscriptionModelOptionsSchema
->;
+import { falTranscriptionModelOptionsSchema } from './fal-transcription-model-options';
+import type { FalTranscriptionModelId } from './fal-transcription-options';
+import type { FalTranscriptionAPITypes } from './fal-api-types';
 
 interface FalTranscriptionModelConfig extends FalConfig {
   _internal?: {
@@ -74,6 +36,20 @@ export class FalTranscriptionModel implements TranscriptionModelV4 {
     return this.config.provider;
   }
 
+  static [WORKFLOW_SERIALIZE](model: FalTranscriptionModel) {
+    return serializeModelOptions({
+      modelId: model.modelId,
+      config: model.config,
+    });
+  }
+
+  static [WORKFLOW_DESERIALIZE](options: {
+    modelId: FalTranscriptionModelId;
+    config: FalTranscriptionModelConfig;
+  }) {
+    return new FalTranscriptionModel(options.modelId, options.config);
+  }
+
   constructor(
     readonly modelId: FalTranscriptionModelId,
     private readonly config: FalTranscriptionModelConfig,
@@ -138,7 +114,7 @@ export class FalTranscriptionModel implements TranscriptionModelV4 {
         path: `https://queue.fal.run/fal-ai/${this.modelId}`,
         modelId: this.modelId,
       }),
-      headers: combineHeaders(this.config.headers(), options.headers),
+      headers: combineHeaders(this.config.headers?.(), options.headers),
       body: {
         ...body,
         audio_url: audioUrl,
@@ -170,7 +146,7 @@ export class FalTranscriptionModel implements TranscriptionModelV4 {
             path: `https://queue.fal.run/fal-ai/${this.modelId}/requests/${queueResponse.request_id}`,
             modelId: this.modelId,
           }),
-          headers: combineHeaders(this.config.headers(), options.headers),
+          headers: combineHeaders(this.config.headers?.(), options.headers),
           failedResponseHandler: async ({
             requestBodyValues,
             response,

package/src/fal-video-model-options.ts

@@ -0,0 +1,39 @@
+import { lazySchema, zodSchema } from '@ai-sdk/provider-utils';
+import { z } from 'zod/v4';
+
+export type FalVideoModelOptions = {
+  loop?: boolean | null;
+  motionStrength?: number | null;
+  pollIntervalMs?: number | null;
+  pollTimeoutMs?: number | null;
+  resolution?: string | null;
+  negativePrompt?: string | null;
+  promptOptimizer?: boolean | null;
+  [key: string]: unknown; // For passthrough
+};
+
+// Provider options schema for FAL video generation
+export const falVideoModelOptionsSchema = lazySchema(() =>
+  zodSchema(
+    z
+      .object({
+        // Video loop - only for Luma models
+        loop: z.boolean().nullish(),
+
+        // Motion strength (provider-specific)
+        motionStrength: z.number().min(0).max(1).nullish(),
+
+        // Polling configuration
+        pollIntervalMs: z.number().positive().nullish(),
+        pollTimeoutMs: z.number().positive().nullish(),
+
+        // Resolution (model-specific, e.g., '480p', '720p', '1080p')
+        resolution: z.string().nullish(),
+
+        // Model-specific parameters
+        negativePrompt: z.string().nullish(),
+        promptOptimizer: z.boolean().nullish(),
+      })
+      .passthrough(),
+  ),
+);

package/src/fal-video-model.ts

@@ -10,53 +10,18 @@ import {
   createJsonResponseHandler,
   delay,
   getFromApi,
-  lazySchema,
   parseProviderOptions,
   postJsonToApi,
-  zodSchema,
 } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
 import type { FalConfig } from './fal-config';
 import { falErrorDataSchema, falFailedResponseHandler } from './fal-error';
+import {
+  falVideoModelOptionsSchema,
+  type FalVideoModelOptions,
+} from './fal-video-model-options';
 import type { FalVideoModelId } from './fal-video-settings';
 
-export type FalVideoModelOptions = {
-  loop?: boolean | null;
-  motionStrength?: number | null;
-  pollIntervalMs?: number | null;
-  pollTimeoutMs?: number | null;
-  resolution?: string | null;
-  negativePrompt?: string | null;
-  promptOptimizer?: boolean | null;
-  [key: string]: unknown; // For passthrough
-};
-
-// Provider options schema for FAL video generation
-export const falVideoModelOptionsSchema = lazySchema(() =>
-  zodSchema(
-    z
-      .object({
-        // Video loop - only for Luma models
-        loop: z.boolean().nullish(),
-
-        // Motion strength (provider-specific)
-        motionStrength: z.number().min(0).max(1).nullish(),
-
-        // Polling configuration
-        pollIntervalMs: z.number().positive().nullish(),
-        pollTimeoutMs: z.number().positive().nullish(),
-
-        // Resolution (model-specific, e.g., '480p', '720p', '1080p')
-        resolution: z.string().nullish(),
-
-        // Model-specific parameters
-        negativePrompt: z.string().nullish(),
-        promptOptimizer: z.boolean().nullish(),
-      })
-      .passthrough(),
-  ),
-);
-
 interface FalVideoModelConfig extends FalConfig {
   _internal?: {
     currentDate?: () => Date;
@@ -158,7 +123,7 @@ export class FalVideoModel implements Experimental_VideoModelV4 {
         path: `https://queue.fal.run/fal-ai/${this.normalizedModelId}`,
         modelId: this.modelId,
       }),
-      headers: combineHeaders(this.config.headers(), options.headers),
+      headers: combineHeaders(this.config.headers?.(), options.headers),
       body,
       failedResponseHandler: falFailedResponseHandler,
       successfulResponseHandler:
@@ -189,7 +154,7 @@ export class FalVideoModel implements Experimental_VideoModelV4 {
               path: responseUrl,
               modelId: this.modelId,
             }),
-            headers: combineHeaders(this.config.headers(), options.headers),
+            headers: combineHeaders(this.config.headers?.(), options.headers),
             failedResponseHandler: async ({
               response,
               url,

package/src/index.ts

@@ -4,13 +4,13 @@ export type {
   FalImageModelOptions,
   /** @deprecated Use `FalImageModelOptions` instead. */
   FalImageModelOptions as FalImageProviderOptions,
-} from './fal-image-options';
-export type { FalSpeechModelOptions } from './fal-speech-model';
-export type { FalTranscriptionModelOptions } from './fal-transcription-model';
+} from './fal-image-model-options';
+export type { FalSpeechModelOptions } from './fal-speech-model-options';
+export type { FalTranscriptionModelOptions } from './fal-transcription-model-options';
 export type {
   FalVideoModelOptions,
   /** @deprecated Use `FalVideoModelOptions` instead. */
   FalVideoModelOptions as FalVideoProviderOptions,
-} from './fal-video-model';
+} from './fal-video-model-options';
 export type { FalVideoModelId } from './fal-video-settings';
 export { VERSION } from './version';

package/dist/index.d.mts

@@ -1,156 +0,0 @@
-import { ProviderV4, ImageModelV4, TranscriptionModelV4, Experimental_VideoModelV4, SpeechModelV4 } from '@ai-sdk/provider';
-import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
-import { FetchFunction, InferSchema } from '@ai-sdk/provider-utils';
-import { z } from 'zod/v4';
-
-type FalImageModelId = 'fal-ai/aura-sr' | 'fal-ai/bria/background/remove' | 'fal-ai/bria/eraser' | 'fal-ai/bria/product-shot' | 'fal-ai/bria/reimagine' | 'bria/text-to-image/3.2' | 'fal-ai/bria/text-to-image/base' | 'fal-ai/bria/text-to-image/fast' | 'fal-ai/bria/text-to-image/hd' | 'fal-ai/bytedance/dreamina/v3.1/text-to-image' | 'fal-ai/ccsr' | 'fal-ai/clarity-upscaler' | 'fal-ai/creative-upscaler' | 'fal-ai/esrgan' | 'fal-ai/flux-general' | 'fal-ai/flux-general/differential-diffusion' | 'fal-ai/flux-general/image-to-image' | 'fal-ai/flux-general/inpainting' | 'fal-ai/flux-general/rf-inversion' | 'fal-ai/flux-kontext-lora/text-to-image' | 'fal-ai/flux-lora' | 'fal-ai/flux-lora/image-to-image' | 'fal-ai/flux-lora/inpainting' | 'fal-ai/flux-pro/kontext' | 'fal-ai/flux-pro/kontext/max' | 'fal-ai/flux-pro/v1.1' | 'fal-ai/flux-pro/v1.1-ultra' | 'fal-ai/flux-pro/v1.1-ultra-finetuned' | 'fal-ai/flux-pro/v1.1-ultra/redux' | 'fal-ai/flux-pro/v1.1/redux' | 'fal-ai/flux/dev' | 'fal-ai/flux/dev/image-to-image' | 'fal-ai/flux/dev/redux' | 'fal-ai/flux/krea' | 'fal-ai/flux/krea/image-to-image' | 'fal-ai/flux/krea/redux' | 'fal-ai/flux/schnell' | 'fal-ai/flux/schnell/redux' | 'fal-ai/ideogram/character' | 'fal-ai/ideogram/character/edit' | 'fal-ai/ideogram/character/remix' | 'fal-ai/imagen4/preview' | 'fal-ai/luma-photon' | 'fal-ai/luma-photon/flash' | 'fal-ai/object-removal' | 'fal-ai/omnigen-v2' | 'fal-ai/qwen-image' | 'fal-ai/recraft/v3/text-to-image' | 'fal-ai/recraft/v3/image-to-image' | 'fal-ai/sana/sprint' | 'fal-ai/sana/v1.5/4.8b' | 'fal-ai/sana/v1.5/1.6b' | 'fal-ai/sky-raccoon' | 'fal-ai/wan/v2.2-5b/text-to-image' | 'fal-ai/wan/v2.2-a14b/text-to-image' | 'fal-ai/fashn/tryon/v1.6' | (string & {});
-
-type FalTranscriptionModelId = 'whisper' | 'wizper' | (string & {});
-
-type FalSpeechModelId = 'fal-ai/minimax/voice-clone' | 'fal-ai/minimax/voice-design' | 'fal-ai/dia-tts/voice-clone' | 'fal-ai/minimax/speech-02-hd' | 'fal-ai/minimax/speech-02-turbo' | 'fal-ai/dia-tts' | 'resemble-ai/chatterboxhd/text-to-speech' | (string & {});
-
-type FalVideoModelId = 'luma-dream-machine' | 'luma-ray-2' | 'luma-ray-2-flash' | 'minimax-video' | 'minimax-video-01' | 'hunyuan-video' | (string & {});
-
-interface FalProviderSettings {
-    /**
-     * fal.ai API key. Default value is taken from the `FAL_API_KEY` environment
-     * variable, falling back to `FAL_KEY`.
-     */
-    apiKey?: string;
-    /**
-     * Base URL for the API calls.
-     * The default prefix is `https://fal.run`.
-     */
-    baseURL?: string;
-    /**
-     * Custom headers to include in the requests.
-     */
-    headers?: Record<string, string>;
-    /**
-     * Custom fetch implementation. You can use it as a middleware to intercept
-     * requests, or to provide a custom fetch implementation for e.g. testing.
-     */
-    fetch?: FetchFunction;
-}
-interface FalProvider extends ProviderV4 {
-    /**
-     * Creates a model for image generation.
-     */
-    image(modelId: FalImageModelId): ImageModelV4;
-    /**
-     * Creates a model for image generation.
-     */
-    imageModel(modelId: FalImageModelId): ImageModelV4;
-    /**
-     * Creates a model for transcription.
-     */
-    transcription(modelId: FalTranscriptionModelId): TranscriptionModelV4;
-    /**
-     * Creates a model for video generation.
-     */
-    video(modelId: FalVideoModelId): Experimental_VideoModelV4;
-    /**
-     * Creates a model for video generation.
-     */
-    videoModel(modelId: FalVideoModelId): Experimental_VideoModelV4;
-    /**
-     * Creates a model for speech generation.
-     */
-    speech(modelId: FalSpeechModelId): SpeechModelV4;
-    /**
-     * @deprecated Use `embeddingModel` instead.
-     */
-    textEmbeddingModel(modelId: string): never;
-}
-/**
- * Create a fal.ai provider instance.
- */
-declare function createFal(options?: FalProviderSettings): FalProvider;
-/**
- * Default fal.ai provider instance.
- */
-declare const fal: FalProvider;
-
-declare const falImageModelOptionsSchema: _ai_sdk_provider_utils.LazySchema<Record<string, unknown>>;
-type FalImageModelOptions = InferSchema<typeof falImageModelOptionsSchema>;
-
-declare const falSpeechModelOptionsSchema: z.ZodObject<{
-    voice_setting: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        speed: z.ZodOptional<z.ZodOptional<z.ZodNullable<z.ZodNumber>>>;
-        vol: z.ZodOptional<z.ZodOptional<z.ZodNullable<z.ZodNumber>>>;
-        voice_id: z.ZodOptional<z.ZodOptional<z.ZodNullable<z.ZodString>>>;
-        pitch: z.ZodOptional<z.ZodOptional<z.ZodNullable<z.ZodNumber>>>;
-        english_normalization: z.ZodOptional<z.ZodOptional<z.ZodNullable<z.ZodBoolean>>>;
-        emotion: z.ZodOptional<z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-            happy: "happy";
-            sad: "sad";
-            angry: "angry";
-            fearful: "fearful";
-            disgusted: "disgusted";
-            surprised: "surprised";
-            neutral: "neutral";
-        }>>>>;
-    }, z.core.$strip>>>;
-    audio_setting: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodUnknown>>>;
-    language_boost: z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-        Chinese: "Chinese";
-        "Chinese,Yue": "Chinese,Yue";
-        English: "English";
-        Arabic: "Arabic";
-        Russian: "Russian";
-        Spanish: "Spanish";
-        French: "French";
-        Portuguese: "Portuguese";
-        German: "German";
-        Turkish: "Turkish";
-        Dutch: "Dutch";
-        Ukrainian: "Ukrainian";
-        Vietnamese: "Vietnamese";
-        Indonesian: "Indonesian";
-        Japanese: "Japanese";
-        Italian: "Italian";
-        Korean: "Korean";
-        Thai: "Thai";
-        Polish: "Polish";
-        Romanian: "Romanian";
-        Greek: "Greek";
-        Czech: "Czech";
-        Finnish: "Finnish";
-        Hindi: "Hindi";
-        auto: "auto";
-    }>>>;
-    pronunciation_dict: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodString>>>;
-}, z.core.$loose>;
-type FalSpeechModelOptions = z.infer<typeof falSpeechModelOptionsSchema>;
-
-declare const falTranscriptionModelOptionsSchema: z.ZodObject<{
-    language: z.ZodDefault<z.ZodOptional<z.ZodNullable<z.ZodUnion<readonly [z.ZodEnum<{
-        en: "en";
-    }>, z.ZodString]>>>>;
-    diarize: z.ZodDefault<z.ZodOptional<z.ZodNullable<z.ZodBoolean>>>;
-    chunkLevel: z.ZodDefault<z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-        segment: "segment";
-        word: "word";
-    }>>>>;
-    version: z.ZodDefault<z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-        3: "3";
-    }>>>>;
-    batchSize: z.ZodDefault<z.ZodOptional<z.ZodNullable<z.ZodNumber>>>;
-    numSpeakers: z.ZodOptional<z.ZodNullable<z.ZodNullable<z.ZodNumber>>>;
-}, z.core.$strip>;
-type FalTranscriptionModelOptions = z.infer<typeof falTranscriptionModelOptionsSchema>;
-
-type FalVideoModelOptions = {
-    loop?: boolean | null;
-    motionStrength?: number | null;
-    pollIntervalMs?: number | null;
-    pollTimeoutMs?: number | null;
-    resolution?: string | null;
-    negativePrompt?: string | null;
-    promptOptimizer?: boolean | null;
-    [key: string]: unknown;
-};
-
-declare const VERSION: string;
-
-export { type FalImageModelOptions, type FalImageModelOptions as FalImageProviderOptions, type FalProvider, type FalProviderSettings, type FalSpeechModelOptions, type FalTranscriptionModelOptions, type FalVideoModelId, type FalVideoModelOptions, type FalVideoModelOptions as FalVideoProviderOptions, VERSION, createFal, fal };