@apicity/kie 0.1.0-alpha.0

package/dist/src/zod.js

@@ -0,0 +1,866 @@
+import { z } from "zod";
+// ---------------------------------------------------------------------------
+// Enums / named union types
+// ---------------------------------------------------------------------------
+export const KieMediaModelSchema = z.enum([
+    "kling-3.0/video",
+    "kling-3.0/motion-control",
+    "grok-imagine/text-to-image",
+    "grok-imagine/image-to-image",
+    "grok-imagine/text-to-video",
+    "grok-imagine/image-to-video",
+    "nano-banana-pro",
+    "nano-banana-2",
+    "gpt-image/1.5-image-to-image",
+    "gpt-image-2-image-to-image",
+    "gpt-image-2-text-to-image",
+    "seedream/5-lite-image-to-image",
+    "seedream/5-lite-text-to-image",
+    "grok-imagine/extend",
+    "grok-imagine/upscale",
+    "qwen2/text-to-image",
+    "qwen2/image-edit",
+    "bytedance/seedance-2-fast",
+    "bytedance/seedance-2",
+    "wan/2-7-image-to-video",
+    "wan/2-7-text-to-video",
+    "wan/2-7-r2v",
+    "wan/2-7-videoedit",
+    "wan/2-7-image",
+    "wan/2-7-image-pro",
+    "happyhorse/text-to-video",
+    "happyhorse/image-to-video",
+    "happyhorse/reference-to-video",
+    "happyhorse/video-edit",
+    "sora-watermark-remover",
+]);
+export const MediaTypeSchema = z.enum([
+    "image",
+    "video",
+    "audio",
+    "transcription",
+]);
+export const KlingDurationSchema = z.enum([
+    "3",
+    "4",
+    "5",
+    "6",
+    "7",
+    "8",
+    "9",
+    "10",
+    "11",
+    "12",
+    "13",
+    "14",
+    "15",
+]);
+export const KlingAspectRatioSchema = z.enum(["16:9", "9:16", "1:1"]);
+export const KlingModeSchema = z.enum(["std", "pro", "4K"]);
+export const GrokImagineModeSchema = z.enum(["fun", "normal", "spicy"]);
+export const GrokImagineDurationSchema = z.enum(["6", "10"]);
+export const GrokImagineResolutionSchema = z.enum(["480p", "720p"]);
+export const NanoBananaResolutionSchema = z.enum(["1K", "2K", "4K"]);
+export const NanoBananaOutputFormatSchema = z.enum(["png", "jpg"]);
+export const GptImageQualitySchema = z.enum(["medium", "high"]);
+export const Qwen2ImageSizeSchema = z.enum([
+    "square",
+    "square_hd",
+    "portrait_4_3",
+    "portrait_16_9",
+    "landscape_4_3",
+    "landscape_16_9",
+]);
+export const Qwen2AccelerationSchema = z.enum(["none", "regular", "high"]);
+export const Wan27ResolutionSchema = z.enum(["720p", "1080p"]);
+export const Wan27AspectRatioSchema = z.enum([
+    "16:9",
+    "9:16",
+    "1:1",
+    "4:3",
+    "3:4",
+]);
+export const Wan27AudioSettingSchema = z.enum(["auto", "origin"]);
+export const Wan27ImageResolutionSchema = z.enum(["1K", "2K", "4K"]);
+export const Wan27ImageAspectRatioSchema = z.enum([
+    "1:1",
+    "16:9",
+    "4:3",
+    "21:9",
+    "3:4",
+    "9:16",
+    "8:1",
+    "1:8",
+]);
+export const HappyHorseResolutionSchema = z.enum(["720p", "1080p"]);
+export const HappyHorseAspectRatioSchema = z.enum([
+    "16:9",
+    "9:16",
+    "1:1",
+    "4:3",
+    "3:4",
+]);
+export const HappyHorseAudioSettingSchema = z.enum(["auto", "origin"]);
+// ---------------------------------------------------------------------------
+// Sub-schemas (composable building blocks)
+// ---------------------------------------------------------------------------
+export const KlingElementSchema = z.object({
+    name: z.string().min(1),
+    description: z.string().min(1),
+    element_input_urls: z.array(z.string()).optional(),
+    element_input_video_urls: z.array(z.string()).optional(),
+});
+export const MultiShotPromptSchema = z.object({
+    prompt: z.string().min(1).max(500),
+    duration: z.number().int().min(1).max(12),
+});
+export const Wan27ImageColorPaletteSchema = z.object({
+    hex: z.string().min(1),
+    ratio: z.string().min(1),
+});
+// ---------------------------------------------------------------------------
+// Media request schemas
+// ---------------------------------------------------------------------------
+export const KlingVideoRequestSchema = z.object({
+    model: z.literal("kling-3.0/video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().optional(),
+        image_urls: z.array(z.string()).optional(),
+        sound: z.boolean(),
+        duration: KlingDurationSchema,
+        aspect_ratio: KlingAspectRatioSchema.optional(),
+        mode: KlingModeSchema,
+        multi_shots: z.boolean(),
+        multi_prompt: z.array(MultiShotPromptSchema).optional(),
+        kling_elements: z.array(KlingElementSchema).max(3).optional(),
+    }),
+});
+export const KlingMotionControlRequestSchema = z.object({
+    model: z.literal("kling-3.0/motion-control"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().optional(),
+        input_urls: z.array(z.string()).min(1).max(1),
+        video_urls: z.array(z.string()).min(1),
+        mode: z.enum(["720p", "1080p"]).optional(),
+        character_orientation: z.enum(["video", "image"]).optional(),
+        background_source: z.enum(["input_video", "input_image"]).optional(),
+    }),
+});
+export const GrokTextToImageRequestSchema = z.object({
+    model: z.literal("grok-imagine/text-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        aspect_ratio: z.enum(["2:3", "3:2", "1:1", "16:9", "9:16"]).default("16:9"),
+        nsfw_checker: z.boolean().default(false),
+        enable_pro: z.boolean().optional(),
+    }),
+});
+export const Qwen2TextToImageRequestSchema = z.object({
+    model: z.literal("qwen2/text-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1),
+        image_size: Qwen2ImageSizeSchema.optional(),
+        num_inference_steps: z.number().optional(),
+        seed: z.number().optional(),
+        guidance_scale: z.number().optional(),
+        enable_safety_checker: z.boolean().optional(),
+        output_format: z.enum(["png", "jpeg"]).optional(),
+        negative_prompt: z.string().optional(),
+        acceleration: Qwen2AccelerationSchema.optional(),
+    }),
+});
+export const Qwen2ImageEditRequestSchema = z.object({
+    model: z.literal("qwen2/image-edit"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1),
+        image_url: z.array(z.string().min(1)).min(1).max(3),
+        image_size: z
+            .enum(["1:1", "2:3", "3:2", "3:4", "4:3", "9:16", "16:9", "21:9"])
+            .default("16:9"),
+        output_format: z.enum(["png", "jpeg"]).default("png"),
+        seed: z.number().optional(),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+export const GrokImageToImageRequestSchema = z.object({
+    model: z.literal("grok-imagine/image-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().max(390000).optional(),
+        image_urls: z.tuple([z.string()]),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+export const GrokTextToVideoRequestSchema = z.object({
+    model: z.literal("grok-imagine/text-to-video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        aspect_ratio: z.enum(["2:3", "3:2", "1:1", "16:9", "9:16"]).optional(),
+        mode: GrokImagineModeSchema.optional(),
+        duration: z.number().int().min(6).max(30).optional(),
+        resolution: GrokImagineResolutionSchema.optional(),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+// Spec types image-to-video duration as a string (e.g. "6"), even though the
+// underlying value is a 6–30 second integer. text-to-video uses a number.
+export const GrokImageToVideoDurationSchema = z
+    .string()
+    .regex(/^([6-9]|[12][0-9]|30)$/);
+export const GrokImageToVideoRequestSchema = z.object({
+    model: z.literal("grok-imagine/image-to-video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().max(5000).optional(),
+        image_urls: z.array(z.string()).max(7).optional(),
+        task_id: z.string().max(100).optional(),
+        index: z.number().int().min(0).max(5).optional(),
+        mode: GrokImagineModeSchema.optional(),
+        duration: GrokImageToVideoDurationSchema.optional(),
+        resolution: GrokImagineResolutionSchema.optional(),
+        aspect_ratio: z.enum(["2:3", "3:2", "1:1", "16:9", "9:16"]).optional(),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+export const GrokVideoExtendRequestSchema = z.object({
+    model: z.literal("grok-imagine/extend"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        task_id: z.string().min(1).max(100),
+        prompt: z.string().min(1).max(5000),
+        extend_at: z.string(),
+        extend_times: GrokImagineDurationSchema,
+    }),
+});
+export const GrokVideoUpscaleRequestSchema = z.object({
+    model: z.literal("grok-imagine/upscale"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        task_id: z.string().min(1).max(100),
+    }),
+});
+export const NanoBananaProRequestSchema = z.object({
+    model: z.literal("nano-banana-pro"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1),
+        image_input: z.array(z.string()).max(8).optional(),
+        aspect_ratio: z
+            .enum([
+            "1:1",
+            "2:3",
+            "3:2",
+            "3:4",
+            "4:3",
+            "4:5",
+            "5:4",
+            "9:16",
+            "16:9",
+            "21:9",
+            "auto",
+        ])
+            .default("16:9"),
+        resolution: NanoBananaResolutionSchema.default("2K"),
+        output_format: NanoBananaOutputFormatSchema.optional(),
+    }),
+});
+// Inner input schema kept unrefined so callers can walk `.shape` for slot
+// introspection (see Seedance2InputSchema for full rationale).
+export const Seedance2FastInputSchema = z.object({
+    prompt: z.string().min(3).max(20000),
+    first_frame_url: z.string().optional(),
+    last_frame_url: z.string().optional(),
+    reference_image_urls: z.array(z.string()).max(9).optional(),
+    reference_video_urls: z.array(z.string()).max(3).optional(),
+    reference_audio_urls: z.array(z.string()).max(3).optional(),
+    /** @deprecated */
+    return_last_frame: z.boolean().optional(),
+    generate_audio: z.boolean().optional(),
+    resolution: z.enum(["480p", "720p"]).optional(),
+    aspect_ratio: z
+        .enum(["1:1", "4:3", "3:4", "16:9", "9:16", "21:9", "adaptive"])
+        .optional(),
+    duration: z.number().int().min(4).max(15).default(5),
+    web_search: z.boolean().optional(),
+    nsfw_checker: z.boolean().default(false),
+});
+const Seedance2FastRequestObjectSchema = z.object({
+    model: z.literal("bytedance/seedance-2-fast"),
+    callBackUrl: z.string().optional(),
+    input: Seedance2FastInputSchema,
+});
+// Mirrors the bytedance/seedance-2 mutual-exclusion rule — fast variant
+// shares the input shape and the same documented constraint that
+// first/last frames and multimodal references are mutually exclusive.
+export const Seedance2FastRequestSchema = Seedance2FastRequestObjectSchema.refine((v) => {
+    const hasReference = (v.input.reference_image_urls?.length ?? 0) > 0 ||
+        (v.input.reference_video_urls?.length ?? 0) > 0 ||
+        (v.input.reference_audio_urls?.length ?? 0) > 0;
+    const hasFrame = Boolean(v.input.first_frame_url) || Boolean(v.input.last_frame_url);
+    return !(hasReference && hasFrame);
+}, {
+    message: "bytedance/seedance-2-fast does not accept reference_image_urls, reference_video_urls, or reference_audio_urls combined with first_frame_url or last_frame_url (these scenarios are mutually exclusive)",
+    path: ["input", "reference_image_urls"],
+});
+// Inner input schema kept unrefined so callers (e.g. clipfirst) can walk
+// `.shape` for slot-constraint introspection — wrapping the request in
+// `.refine()` below turns it into ZodEffects and hides `.shape`.
+export const Seedance2InputSchema = z.object({
+    prompt: z.string().min(3).max(20000),
+    first_frame_url: z.string().optional(),
+    last_frame_url: z.string().optional(),
+    reference_image_urls: z.array(z.string()).max(9).optional(),
+    reference_video_urls: z.array(z.string()).max(3).optional(),
+    reference_audio_urls: z.array(z.string()).max(3).optional(),
+    /** @deprecated */
+    return_last_frame: z.boolean().optional(),
+    generate_audio: z.boolean().optional(),
+    resolution: z.enum(["480p", "720p", "1080p"]).optional(),
+    aspect_ratio: z
+        .enum(["1:1", "4:3", "3:4", "16:9", "9:16", "21:9", "adaptive"])
+        .optional(),
+    duration: z.number().int().min(4).max(15).default(5),
+    web_search: z.boolean().optional(),
+    nsfw_checker: z.boolean().default(false),
+});
+const Seedance2RequestObjectSchema = z.object({
+    model: z.literal("bytedance/seedance-2"),
+    callBackUrl: z.string().optional(),
+    input: Seedance2InputSchema,
+});
+// Per Kie docs, bytedance/seedance-2 has three mutually exclusive scenarios:
+// Image-to-Video (first frame), Image-to-Video (first + last frames), and
+// Multimodal Reference-to-Video (any of reference_image_urls /
+// reference_video_urls / reference_audio_urls). Mixing first/last frames
+// with any reference_* field returns "The reference video and the first
+// and last frames are mutually exclusive, and only one scene can be
+// selected" from createTask. Enforce at the SDK boundary.
+export const Seedance2RequestSchema = Seedance2RequestObjectSchema.refine((v) => {
+    const hasReference = (v.input.reference_image_urls?.length ?? 0) > 0 ||
+        (v.input.reference_video_urls?.length ?? 0) > 0 ||
+        (v.input.reference_audio_urls?.length ?? 0) > 0;
+    const hasFrame = Boolean(v.input.first_frame_url) || Boolean(v.input.last_frame_url);
+    return !(hasReference && hasFrame);
+}, {
+    message: "bytedance/seedance-2 does not accept reference_image_urls, reference_video_urls, or reference_audio_urls combined with first_frame_url or last_frame_url (these scenarios are mutually exclusive)",
+    path: ["input", "reference_image_urls"],
+});
+export const NanoBanana2RequestSchema = z.object({
+    model: z.literal("nano-banana-2"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1),
+        image_input: z.array(z.string()).optional(),
+        aspect_ratio: z
+            .enum([
+            "1:1",
+            "2:3",
+            "3:2",
+            "3:4",
+            "4:3",
+            "4:5",
+            "5:4",
+            "9:16",
+            "16:9",
+            "21:9",
+            "1:4",
+            "1:8",
+            "4:1",
+            "8:1",
+            "auto",
+        ])
+            .default("16:9"),
+        resolution: NanoBananaResolutionSchema.default("2K"),
+        output_format: NanoBananaOutputFormatSchema.optional(),
+        google_search: z.boolean().optional(),
+    }),
+});
+export const GptImageToImageRequestSchema = z.object({
+    model: z.literal("gpt-image/1.5-image-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        input_urls: z.array(z.string()).min(1).max(4),
+        prompt: z.string().min(1),
+        aspect_ratio: z.enum(["1:1", "2:3", "3:2"]).optional(),
+        quality: GptImageQualitySchema.optional(),
+    }),
+});
+export const GptImage2ImageToImageAspectRatioSchema = z.enum([
+    "auto",
+    "1:1",
+    "9:16",
+    "16:9",
+    "4:3",
+    "3:4",
+]);
+export const GptImage2ImageToImageResolutionSchema = z.enum(["1K", "2K", "4K"]);
+export const GptImage2ImageToImageRequestSchema = z.object({
+    model: z.literal("gpt-image-2-image-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(20000),
+        input_urls: z.array(z.string()).min(1).max(16),
+        aspect_ratio: GptImage2ImageToImageAspectRatioSchema.default("16:9"),
+        resolution: GptImage2ImageToImageResolutionSchema.default("2K"),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+export const GptImage2TextToImageAspectRatioSchema = z.enum([
+    "auto",
+    "1:1",
+    "9:16",
+    "16:9",
+    "4:3",
+    "3:4",
+]);
+export const GptImage2TextToImageResolutionSchema = z.enum(["1K", "2K", "4K"]);
+export const GptImage2TextToImageRequestSchema = z.object({
+    model: z.literal("gpt-image-2-text-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(20000),
+        aspect_ratio: GptImage2TextToImageAspectRatioSchema.default("16:9"),
+        resolution: GptImage2TextToImageResolutionSchema.default("2K"),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+// Kie's seedream/5-lite createTask rejects requests with `"This field is
+// required"` when `quality` is missing, even though their docs list it as
+// optional with a default. Treat it as required at the SDK boundary so the
+// type system forces callers to pick basic/high.
+export const SeedreamImageToImageRequestSchema = z.object({
+    model: z.literal("seedream/5-lite-image-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        image_urls: z.array(z.string()).min(1).max(14),
+        prompt: z.string().min(3).max(3000),
+        aspect_ratio: z
+            .enum(["1:1", "4:3", "3:4", "16:9", "9:16", "2:3", "3:2", "21:9"])
+            .default("16:9"),
+        quality: z.enum(["basic", "high"]),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+export const SeedreamTextToImageRequestSchema = z.object({
+    model: z.literal("seedream/5-lite-text-to-image"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(3).max(3000),
+        aspect_ratio: z
+            .enum(["1:1", "4:3", "3:4", "16:9", "9:16", "2:3", "3:2", "21:9"])
+            .default("16:9"),
+        quality: z.enum(["basic", "high"]),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+export const SoraWatermarkRequestSchema = z.object({
+    model: z.literal("sora-watermark-remover"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        video_url: z.string().min(1),
+        upload_method: z.enum(["s3", "oss"]).optional(),
+    }),
+});
+export const HappyHorseTextToVideoRequestSchema = z.object({
+    model: z.literal("happyhorse/text-to-video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        resolution: HappyHorseResolutionSchema.optional(),
+        aspect_ratio: HappyHorseAspectRatioSchema.optional(),
+        duration: z.number().int().min(3).max(15).optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+    }),
+});
+export const HappyHorseImageToVideoRequestSchema = z.object({
+    model: z.literal("happyhorse/image-to-video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().max(5000).optional(),
+        image_urls: z.array(z.string()).min(1).max(1),
+        resolution: HappyHorseResolutionSchema.optional(),
+        duration: z.number().int().min(3).max(15).optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+    }),
+});
+export const HappyHorseReferenceToVideoRequestSchema = z.object({
+    model: z.literal("happyhorse/reference-to-video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        reference_image: z.array(z.string()).min(1).max(9),
+        resolution: HappyHorseResolutionSchema.optional(),
+        aspect_ratio: HappyHorseAspectRatioSchema.optional(),
+        duration: z.number().int().min(3).max(15).optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+    }),
+});
+export const HappyHorseVideoEditRequestSchema = z.object({
+    model: z.literal("happyhorse/video-edit"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        video_url: z.string().min(1),
+        reference_image: z.array(z.string()).max(5).optional(),
+        resolution: HappyHorseResolutionSchema.optional(),
+        audio_setting: HappyHorseAudioSettingSchema.optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+    }),
+});
+// Refines live on the outer request object (not the `input` sub-object) so
+// that `input.*` fields remain introspectable by downstream tools that walk
+// ZodArray/ZodObject defs (e.g. clipfirst's readSlotConstraints).
+export const Wan27ImageToVideoRequestSchema = z
+    .object({
+    model: z.literal("wan/2-7-image-to-video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        negative_prompt: z.string().max(500).optional(),
+        first_frame_url: z.string().optional(),
+        last_frame_url: z.string().optional(),
+        first_clip_url: z.string().optional(),
+        driving_audio_url: z.string().optional(),
+        resolution: Wan27ResolutionSchema.optional(),
+        duration: z.number().int().min(2).max(15).optional(),
+        prompt_extend: z.boolean().optional(),
+        watermark: z.boolean().optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+        nsfw_checker: z.boolean().default(false),
+    }),
+})
+    .refine((v) => Boolean(v.input.first_frame_url) ||
+    Boolean(v.input.last_frame_url) ||
+    Boolean(v.input.first_clip_url), {
+    message: "wan/2-7-image-to-video requires at least one of first_frame_url, last_frame_url, or first_clip_url",
+    path: ["input", "first_frame_url"],
+})
+    .refine((v) => !(v.input.first_clip_url &&
+    (v.input.first_frame_url || v.input.last_frame_url)), {
+    message: "wan/2-7-image-to-video does not accept first_clip_url combined with first_frame_url or last_frame_url",
+    path: ["input", "first_clip_url"],
+});
+export const Wan27TextToVideoRequestSchema = z.object({
+    model: z.literal("wan/2-7-text-to-video"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        negative_prompt: z.string().max(500).optional(),
+        audio_url: z.string().optional(),
+        resolution: Wan27ResolutionSchema.optional(),
+        ratio: Wan27AspectRatioSchema.optional(),
+        duration: z.number().int().min(2).max(15).optional(),
+        prompt_extend: z.boolean().optional(),
+        watermark: z.boolean().optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+// Per-field max(5) on reference_image/reference_video; the combined-≤5 cap is
+// enforced by callers (would require a wrapper-level refine, which would
+// turn this into ZodEffects and break `.shape` introspection used by
+// clipfirst's slot-constraint readers).
+export const Wan27RefToVideoRequestSchema = z.object({
+    model: z.literal("wan/2-7-r2v"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().min(1).max(5000),
+        negative_prompt: z.string().max(500).optional(),
+        reference_image: z.array(z.string()).max(5).optional(),
+        reference_video: z.array(z.string()).max(5).optional(),
+        first_frame: z.string().optional(),
+        reference_voice: z.string().optional(),
+        resolution: Wan27ResolutionSchema.optional(),
+        aspect_ratio: Wan27AspectRatioSchema.optional(),
+        duration: z.number().int().min(2).max(10).optional(),
+        prompt_extend: z.boolean().optional(),
+        watermark: z.boolean().optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+        nsfw_checker: z.boolean().default(false),
+    }),
+});
+export const Wan27VideoEditRequestSchema = z
+    .object({
+    model: z.literal("wan/2-7-videoedit"),
+    callBackUrl: z.string().optional(),
+    input: z.object({
+        prompt: z.string().max(5000).optional(),
+        negative_prompt: z.string().max(500).optional(),
+        video_url: z.string().min(1),
+        reference_image: z.string().optional(),
+        resolution: Wan27ResolutionSchema.optional(),
+        aspect_ratio: Wan27AspectRatioSchema.optional(),
+        duration: z.number().int().optional(),
+        audio_setting: Wan27AudioSettingSchema.optional(),
+        prompt_extend: z.boolean().optional(),
+        watermark: z.boolean().optional(),
+        seed: z.number().int().min(0).max(2147483647).optional(),
+        nsfw_checker: z.boolean().default(false),
+    }),
+})
+    .refine((v) => {
+    const d = v.input.duration;
+    return d === undefined || d === 0 || (d >= 2 && d <= 10);
+}, {
+    message: "wan/2-7-videoedit duration must be 0 (full input) or between 2 and 10 seconds",
+    path: ["input", "duration"],
+});
+const Wan27ImageInputShape = {
+    prompt: z.string().min(1).max(5000),
+    input_urls: z.array(z.string()).max(9).optional(),
+    aspect_ratio: Wan27ImageAspectRatioSchema.default("16:9"),
+    enable_sequential: z.boolean().optional(),
+    n: z.number().int().min(1).max(12).optional(),
+    resolution: Wan27ImageResolutionSchema.default("2K"),
+    thinking_mode: z.boolean().optional(),
+    color_palette: z
+        .array(Wan27ImageColorPaletteSchema)
+        .min(3)
+        .max(10)
+        .optional(),
+    bbox_list: z.array(z.array(z.array(z.number()).length(4)).max(2)).optional(),
+    watermark: z.boolean().optional(),
+    seed: z.number().int().min(0).max(2147483647).optional(),
+    nsfw_checker: z.boolean().default(false),
+};
+const wan27ImageAspectRatioRequiresResolution = (v) => v.input.aspect_ratio === undefined || v.input.resolution !== undefined;
+const wan27ImageAspectRatioRefinement = {
+    message: "aspect_ratio requires resolution",
+    path: ["input", "resolution"],
+};
+const wan27Image4KRequiresNonSequentialTextToImage = (v) => {
+    if (v.input.resolution !== "4K")
+        return true;
+    if (v.input.enable_sequential === true)
+        return false;
+    const urls = v.input.input_urls;
+    if (Array.isArray(urls) && urls.length > 0)
+        return false;
+    return true;
+};
+const wan27Image4KRefinement = {
+    message: "resolution 4K is only supported for non-sequential text-to-image",
+    path: ["input", "resolution"],
+};
+export const Wan27ImageInputSchema = z.object(Wan27ImageInputShape);
+const Wan27ImageRequestObjectSchema = z.object({
+    model: z.literal("wan/2-7-image"),
+    callBackUrl: z.string().optional(),
+    input: Wan27ImageInputSchema,
+});
+const Wan27ImageProRequestObjectSchema = z.object({
+    model: z.literal("wan/2-7-image-pro"),
+    callBackUrl: z.string().optional(),
+    input: Wan27ImageInputSchema,
+});
+export const Wan27ImageRequestSchema = Wan27ImageRequestObjectSchema.refine(wan27ImageAspectRatioRequiresResolution, wan27ImageAspectRatioRefinement).refine(wan27Image4KRequiresNonSequentialTextToImage, wan27Image4KRefinement);
+export const Wan27ImageProRequestSchema = Wan27ImageProRequestObjectSchema.refine(wan27ImageAspectRatioRequiresResolution, wan27ImageAspectRatioRefinement).refine(wan27Image4KRequiresNonSequentialTextToImage, wan27Image4KRefinement);
+// ---------------------------------------------------------------------------
+// Wan 2.7 task result schemas (parsed from KieTaskInfoData.resultJson)
+//
+// kie wraps async task results in a JSON envelope on `resultJson`. Both image
+// and video Wan 2.7 endpoints return the same shape: an array of result URLs
+// (image URLs for image jobs; one video URL for video jobs) plus an optional
+// passthrough `resultObject` for endpoint-specific metadata. Consumers should
+// `JSON.parse(data.resultJson)` then validate with these.
+// ---------------------------------------------------------------------------
+export const Wan27TaskResultJsonSchema = z.object({
+    resultUrls: z.array(z.string().url()).min(1),
+    resultObject: z.record(z.string(), z.unknown()).optional(),
+});
+export const Wan27VideoResultSchema = Wan27TaskResultJsonSchema;
+export const Wan27ImageResultSchema = Wan27TaskResultJsonSchema;
+// ---------------------------------------------------------------------------
+// Upload schemas
+// ---------------------------------------------------------------------------
+const blobSchema = z.instanceof(Blob);
+export const UploadMediaRequestSchema = z.object({
+    file: blobSchema,
+    filename: z.string().min(1),
+    uploadPath: z.string().min(1),
+    fileName: z.string().optional(),
+    mimeType: z.string().optional(),
+});
+export const FileUrlUploadRequestSchema = z.object({
+    fileUrl: z.string().min(1),
+    uploadPath: z.string().min(1),
+    fileName: z.string().optional(),
+});
+export const FileBase64UploadRequestSchema = z.object({
+    base64Data: z.string().min(1),
+    uploadPath: z.string().min(1),
+    fileName: z.string().optional(),
+    mimeType: z.string().optional(),
+});
+// ---------------------------------------------------------------------------
+// Download URL
+// ---------------------------------------------------------------------------
+export const DownloadUrlRequestSchema = z.object({
+    url: z.string().min(1),
+});
+// ---------------------------------------------------------------------------
+// Options
+// ---------------------------------------------------------------------------
+export const KieOptionsSchema = z.object({
+    apiKey: z.string().min(1),
+    baseURL: z.string().url().optional(),
+    uploadBaseURL: z.string().url().optional(),
+    timeout: z.number().int().positive().optional(),
+    fetch: z
+        .custom()
+        .optional(),
+});
+// ---------------------------------------------------------------------------
+// CreateTask request (alias for MediaGenerationRequest — what the createTask
+// endpoint actually receives)
+// ---------------------------------------------------------------------------
+export const CreateTaskRequestSchema = z.object({
+    model: KieMediaModelSchema,
+    callBackUrl: z.string().optional(),
+    input: z.record(z.string(), z.unknown()),
+});
+// ---------------------------------------------------------------------------
+// Sub-provider schemas: Veo
+// ---------------------------------------------------------------------------
+export const VeoGenerateRequestSchema = z.object({
+    prompt: z.string().min(1),
+    model: z.enum(["veo3", "veo3_fast"]).optional(),
+    aspectRatio: z.enum(["16:9", "9:16", "Auto"]).optional(),
+    generationType: z
+        .enum([
+        "TEXT_2_VIDEO",
+        "REFERENCE_2_VIDEO",
+        "FIRST_AND_LAST_FRAMES_2_VIDEO",
+    ])
+        .optional(),
+    imageUrls: z.array(z.string()).optional(),
+    seeds: z.number().optional(),
+    watermark: z.string().optional(),
+    enableTranslation: z.boolean().optional(),
+});
+export const VeoExtendRequestSchema = z.object({
+    taskId: z.string().min(1),
+    prompt: z.string().min(1),
+    model: z.enum(["fast", "quality"]).optional(),
+    seeds: z.number().optional(),
+    watermark: z.string().optional(),
+});
+// ---------------------------------------------------------------------------
+// Sub-provider schemas: Suno
+// ---------------------------------------------------------------------------
+export const SunoGenerateRequestSchema = z.object({
+    prompt: z.string().min(1),
+    model: z.enum(["V3_5", "V4", "V4_5", "V4_5PLUS", "V4_5ALL", "V5", "V5_5"]),
+    instrumental: z.boolean(),
+    customMode: z.boolean(),
+    callBackUrl: z.string().min(1),
+    style: z.string().optional(),
+    negativeTags: z.string().optional(),
+    title: z.string().optional(),
+    vocalGender: z.enum(["m", "f"]).optional(),
+    styleWeight: z.number().min(0).max(1).optional(),
+    weirdnessConstraint: z.number().min(0).max(1).optional(),
+    audioWeight: z.number().min(0).max(1).optional(),
+    personaId: z.string().optional(),
+});
+// ---------------------------------------------------------------------------
+// Sub-provider schemas: Chat (GPT-5.5 / GPT-5.2 via Kie)
+// ---------------------------------------------------------------------------
+export const KieChatContentPartSchema = z.object({
+    type: z.enum(["text", "image_url"]),
+    text: z.string().optional(),
+    image_url: z.object({ url: z.string() }).optional(),
+});
+export const KieChatMessageSchema = z.object({
+    role: z.enum(["user", "assistant", "system"]),
+    content: z.union([z.string(), z.array(KieChatContentPartSchema)]),
+});
+export const KieChatRequestSchema = z.object({
+    model: z.string(),
+    messages: z.array(KieChatMessageSchema),
+    temperature: z.number().min(0).max(2).optional(),
+    max_tokens: z.number().int().positive().optional(),
+    stream: z.boolean().optional(),
+    response_format: z
+        .object({
+        type: z.enum(["text", "json_object", "json_schema"]),
+        json_schema: z.record(z.string(), z.unknown()).optional(),
+    })
+        .optional(),
+});
+// ---------------------------------------------------------------------------
+// Sub-provider schemas: Claude (via Kie)
+// ---------------------------------------------------------------------------
+export const KieClaudeToolInputSchemaSchema = z.object({
+    type: z.string(),
+    properties: z.record(z.string(), z.unknown()).optional(),
+    required: z.array(z.string()).optional(),
+});
+export const KieClaudeToolSchema = z.object({
+    name: z.string(),
+    description: z.string(),
+    input_schema: KieClaudeToolInputSchemaSchema,
+});
+export const KieClaudeContentPartSchema = z
+    .object({
+    type: z.string(),
+    text: z.string().optional(),
+})
+    .passthrough();
+export const KieClaudeMessageSchema = z.object({
+    role: z.enum(["user", "assistant"]),
+    content: z.union([z.string(), z.array(KieClaudeContentPartSchema)]),
+});
+export const KieClaudeRequestSchema = z.object({
+    model: z.enum(["claude-sonnet-4-6", "claude-haiku-4-5"]),
+    messages: z.array(KieClaudeMessageSchema),
+    tools: z.array(KieClaudeToolSchema).optional(),
+    thinkingFlag: z.boolean().optional(),
+    stream: z.boolean().optional(),
+});
+// ---------------------------------------------------------------------------
+// Media generation request (discriminated union on model)
+// ---------------------------------------------------------------------------
+// Plain union (not discriminatedUnion) so individual members can be refined —
+// discriminatedUnion requires ZodObject members, but `.refine()` wraps an
+// object in ZodEffects. Parsing cost is slightly higher (tries each member)
+// but accepted for the added input-contract validation.
+export const MediaGenerationRequestSchema = z.union([
+    KlingVideoRequestSchema,
+    KlingMotionControlRequestSchema,
+    GrokTextToImageRequestSchema,
+    GrokImageToImageRequestSchema,
+    GrokTextToVideoRequestSchema,
+    GrokImageToVideoRequestSchema,
+    GrokVideoExtendRequestSchema,
+    GrokVideoUpscaleRequestSchema,
+    NanoBananaProRequestSchema,
+    NanoBanana2RequestSchema,
+    GptImageToImageRequestSchema,
+    GptImage2ImageToImageRequestSchema,
+    GptImage2TextToImageRequestSchema,
+    SeedreamImageToImageRequestSchema,
+    SeedreamTextToImageRequestSchema,
+    Qwen2TextToImageRequestSchema,
+    Qwen2ImageEditRequestSchema,
+    Seedance2FastRequestSchema,
+    Seedance2RequestSchema,
+    Wan27ImageToVideoRequestSchema,
+    Wan27TextToVideoRequestSchema,
+    Wan27RefToVideoRequestSchema,
+    Wan27VideoEditRequestSchema,
+    Wan27ImageRequestSchema,
+    Wan27ImageProRequestSchema,
+    HappyHorseTextToVideoRequestSchema,
+    HappyHorseImageToVideoRequestSchema,
+    HappyHorseReferenceToVideoRequestSchema,
+    HappyHorseVideoEditRequestSchema,
+    SoraWatermarkRequestSchema,
+]);
+//# sourceMappingURL=zod.js.map