makefx 1.6.6 → 1.6.8

package/README.md

@@ -61,4 +61,4 @@ IDs, or runs. Retired commands and aliases are not redirected.
 
 Run `makefx --help` or `makefx <noun> --help` for current options.
 
-Version: 1.6.6
+Version: 1.6.8

package/makefx.mjs

@@ -4442,7 +4442,7 @@ require_websocket_server();
 var wrapper_default = import_websocket.default;
 //#endregion
 //#region src/cli/version.ts
-var CLI_VERSION = "1.6.6+b1c0c6d222d7";
+var CLI_VERSION = "1.6.8+06570e7d00c2";
 var CLI_VERSION_HEADER = "X-MakeFX-CLI-Version";
 function cliVersionHeaders() {
 	return {
@@ -4461,6 +4461,7 @@ function cliVersionHeaders() {
 var GENERATION_REQUEST_TIMEOUT_MS = 3e5;
 var VIDEO_GENERATION_REQUEST_TIMEOUT_MS = 72e4;
 var VARIANT_SYNC_POLL_INTERVAL_MS = 15e3;
+var GENERATION_RECONNECT_RETRY_MS = 1e3;
 function getGenerationRequestTimeoutMs(mediaKind) {
 	return mediaKind === "video" ? VIDEO_GENERATION_REQUEST_TIMEOUT_MS : GENERATION_REQUEST_TIMEOUT_MS;
 }
@@ -4493,6 +4494,8 @@ var WebSocketClient = class WebSocketClient {
 	env;
 	spaceId;
 	connectionLoggingEnabled = false;
+	disconnectRequested = false;
+	generationReconnect = null;
 	chatHandlers = /* @__PURE__ */ new Map();
 	generateHandlers = /* @__PURE__ */ new Map();
 	earlyTerminalVariants = /* @__PURE__ */ new Map();
@@ -4535,6 +4538,7 @@ var WebSocketClient = class WebSocketClient {
 	* Connect to the WebSocket endpoint
 	*/
 	async connect() {
+		this.disconnectRequested = false;
 		return new Promise((resolve, reject) => {
 			const protocol = this.baseUrl.startsWith("https") ? "wss" : "ws";
 			const url = `${protocol}://${this.baseUrl.replace(/^https?:\/\//, "")}/api/spaces/${this.spaceId}/ws`;
@@ -4563,7 +4567,7 @@ var WebSocketClient = class WebSocketClient {
 			});
 			this.ws.on("close", (code, reason) => {
 				if (this.connectionLoggingEnabled) console.log(`[WebSocketClient] Disconnected: ${code} - ${reason}`);
-				this.ws = null;
+				this.handleSocketClose();
 			});
 		});
 	}
@@ -4571,11 +4575,35 @@ var WebSocketClient = class WebSocketClient {
 	* Disconnect from the WebSocket
 	*/
 	disconnect() {
+		this.disconnectRequested = true;
 		if (this.ws) {
 			this.ws.close();
 			this.ws = null;
 		}
 	}
+	handleSocketClose() {
+		this.ws = null;
+		if (!this.disconnectRequested && this.hasPendingGenerationWaits()) this.reconnectPendingGeneration();
+	}
+	hasPendingGenerationWaits() {
+		return this.generateHandlers.size > 0 || this.variantCompletionHandlers.size > 0;
+	}
+	reconnectPendingGeneration() {
+		if (this.generationReconnect || this.disconnectRequested || !this.hasPendingGenerationWaits()) return;
+		this.generationReconnect = this.runGenerationReconnect().finally(() => {
+			this.generationReconnect = null;
+		});
+	}
+	async runGenerationReconnect() {
+		while (!this.disconnectRequested && this.hasPendingGenerationWaits()) try {
+			await this.connect();
+			if (this.hasPendingGenerationWaits()) this.requestSync();
+			return;
+		} catch {
+			if (this.disconnectRequested || !this.hasPendingGenerationWaits()) return;
+			await new Promise((resolve) => setTimeout(resolve, GENERATION_RECONNECT_RETRY_MS));
+		}
+	}
 	/**
 	* Set error handler
 	*/
@@ -5172,6 +5200,10 @@ var WebSocketClient = class WebSocketClient {
 				videoTier: params.videoTier,
 				seedanceDuration: params.seedanceDuration,
 				seedanceBitrateMode: params.seedanceBitrateMode,
+				wan3Duration: params.wan3Duration,
+				wan3Seed: params.wan3Seed,
+				wan3PromptExpansion: params.wan3PromptExpansion,
+				wan3Thinking: params.wan3Thinking,
 				provider: params.provider,
 				avatarResolution: params.avatarResolution,
 				avatarMode: params.avatarMode,
@@ -6050,7 +6082,7 @@ var SEEDANCE_25_DURATIONS = [
 	30
 ];
 var SEEDANCE_2_BITRATE_MODES = ["standard", "high"];
-var MB = 1024 * 1024;
+var MB$1 = 1024 * 1024;
 var IMAGE_REFERENCE_RULE = {
 	mediaKind: "image",
 	minCount: 0,
@@ -6061,7 +6093,7 @@ var IMAGE_REFERENCE_RULE = {
 		"image/png",
 		"image/webp"
 	],
-	maxBytesPerFile: 30 * MB
+	maxBytesPerFile: 30 * MB$1
 };
 var VIDEO_REFERENCE_RULE = {
 	mediaKind: "video",
@@ -6069,8 +6101,8 @@ var VIDEO_REFERENCE_RULE = {
 	maxCount: 3,
 	promptLabel: "@VideoN",
 	acceptedMimeTypes: ["video/mp4", "video/quicktime"],
-	maxBytesPerFile: 50 * MB,
-	combinedMaxBytes: 50 * MB,
+	maxBytesPerFile: 50 * MB$1,
+	combinedMaxBytes: 50 * MB$1,
 	combinedDurationSeconds: {
 		min: 2,
 		max: 15
@@ -6090,7 +6122,7 @@ var AUDIO_REFERENCE_RULE = {
 		"audio/wav",
 		"audio/x-wav"
 	],
-	maxBytesPerFile: 15 * MB,
+	maxBytesPerFile: 15 * MB$1,
 	combinedDurationSeconds: { max: 15 }
 };
 function profile(mode, tier, endpointId) {
@@ -6159,7 +6191,7 @@ var SEEDANCE_25_IMAGE_REFERENCE_RULE = {
 		"image/webp",
 		"image/gif"
 	],
-	maxBytesPerFile: 30 * MB
+	maxBytesPerFile: 30 * MB$1
 };
 var SEEDANCE_25_FRAME_IMAGE_REFERENCE_RULE = {
 	...SEEDANCE_25_IMAGE_REFERENCE_RULE,
@@ -6177,7 +6209,7 @@ var SEEDANCE_25_VIDEO_REFERENCE_RULE = {
 	maxCount: 10,
 	promptLabel: "@VideoN",
 	acceptedMimeTypes: ["video/mp4", "video/quicktime"],
-	maxBytesPerFile: 50 * MB,
+	maxBytesPerFile: 50 * MB$1,
 	combinedDurationSeconds: { max: 30.2 },
 	perFileDurationSeconds: {
 		min: 1.8,
@@ -6202,7 +6234,7 @@ var SEEDANCE_25_AUDIO_REFERENCE_RULE = {
 		"audio/wav",
 		"audio/x-wav"
 	],
-	maxBytesPerFile: 15 * MB,
+	maxBytesPerFile: 15 * MB$1,
 	combinedDurationSeconds: { max: 30.2 },
 	perFileDurationSeconds: {
 		min: 1.8,
@@ -6343,6 +6375,158 @@ function getSeedance2CapabilityByEndpoint(endpointId) {
 	return SEEDANCE_2_SELECTIONS.map((selection) => SEEDANCE_2_CAPABILITIES[selection]).find((capability) => capability.endpointId === endpointId);
 }
 //#endregion
+//#region src/shared/wan3Capabilities.ts
+var WAN_3_ASPECT_RATIOS = [
+	"adaptive",
+	"16:9",
+	"4:3",
+	"1:1",
+	"3:4",
+	"9:16"
+];
+var WAN_3_RESOLUTIONS = [
+	"480p",
+	"720p",
+	"1080p"
+];
+var WAN_3_DURATIONS = [
+	"auto",
+	2,
+	3,
+	4,
+	5,
+	6,
+	7,
+	8,
+	9,
+	10,
+	11,
+	12,
+	13,
+	14,
+	15,
+	16,
+	17,
+	18,
+	19,
+	20,
+	21,
+	22,
+	23,
+	24,
+	25,
+	26,
+	27,
+	28,
+	29,
+	30
+];
+var MB = 1024 * 1024;
+var IMAGE_RULE = {
+	mediaKind: "image",
+	minCount: 0,
+	maxCount: 10,
+	promptLabel: "@ImageN",
+	acceptedMimeTypes: [
+		"image/jpeg",
+		"image/png",
+		"image/webp"
+	],
+	maxBytesPerFile: 30 * MB
+};
+var FRAME_RULE = {
+	...IMAGE_RULE,
+	minCount: 1,
+	maxCount: 2
+};
+var VIDEO_RULE = {
+	mediaKind: "video",
+	minCount: 0,
+	maxCount: 5,
+	promptLabel: "@VideoN",
+	acceptedMimeTypes: ["video/mp4", "video/quicktime"],
+	maxBytesPerFile: 50 * MB,
+	combinedDurationSeconds: { max: 15 }
+};
+var AUDIO_RULE = {
+	mediaKind: "audio",
+	minCount: 0,
+	maxCount: 5,
+	promptLabel: "@AudioN",
+	acceptedMimeTypes: [
+		"audio/mpeg",
+		"audio/wav",
+		"audio/x-wav"
+	],
+	maxBytesPerFile: 15 * MB,
+	combinedDurationSeconds: { max: 15 }
+};
+function capability(mode, endpointId) {
+	const selection = `wan-3-${mode}`;
+	const references = mode === "frame" ? [FRAME_RULE] : mode === "reference" ? [
+		IMAGE_RULE,
+		VIDEO_RULE,
+		AUDIO_RULE
+	] : [];
+	return {
+		selection,
+		generatorId: `video/${selection}`,
+		endpointId,
+		label: `WAN 3.0 ${mode === "text" ? "Text" : mode === "frame" ? "Frames" : "References"}`,
+		mode,
+		resolutions: WAN_3_RESOLUTIONS,
+		aspectRatios: WAN_3_ASPECT_RATIOS,
+		durations: WAN_3_DURATIONS,
+		defaultResolution: "1080p",
+		defaultAspectRatio: "adaptive",
+		defaultDuration: 5,
+		references,
+		maxReferenceFiles: mode === "frame" ? 2 : mode === "reference" ? 20 : 0
+	};
+}
+var WAN_3_CAPABILITIES = [
+	capability("text", "alibaba/wan-3.0/text-to-video"),
+	capability("frame", "alibaba/wan-3.0/image-to-video"),
+	capability("reference", "alibaba/wan-3.0/reference-to-video")
+];
+var WAN_3_SELECTIONS = WAN_3_CAPABILITIES.map((item) => item.selection);
+WAN_3_CAPABILITIES.map((item) => item.endpointId);
+function getWan3CapabilityBySelection(value) {
+	return WAN_3_CAPABILITIES.find((item) => item.selection === value);
+}
+function getWan3CapabilityByEndpoint(value) {
+	return WAN_3_CAPABILITIES.find((item) => item.endpointId === value);
+}
+function isWan3Duration(value) {
+	return WAN_3_DURATIONS.includes(value);
+}
+function countWan3References(references) {
+	return references.reduce((counts, reference) => {
+		if (reference.mediaKind in counts) counts[reference.mediaKind] += 1;
+		return counts;
+	}, {
+		image: 0,
+		video: 0,
+		audio: 0
+	});
+}
+function getWan3ReferenceError(capability_, references) {
+	const counts = countWan3References(references);
+	const unsupported = references.find((reference) => !capability_.references.some((rule) => rule.mediaKind === reference.mediaKind));
+	if (unsupported) return `${capability_.label} does not accept ${unsupported.mediaKind} references`;
+	if (references.length > capability_.maxReferenceFiles) return `${capability_.label} accepts at most ${capability_.maxReferenceFiles} references`;
+	if (capability_.mode === "reference" && references.length === 0) return "WAN 3.0 reference-to-video requires at least one image, video, or audio reference";
+	for (const rule of capability_.references) {
+		const count = counts[rule.mediaKind];
+		if (count < rule.minCount || count > rule.maxCount) return `${capability_.label} requires ${rule.minCount}-${rule.maxCount} ${rule.mediaKind} references`;
+		if (rule.combinedDurationSeconds) {
+			if (references.filter((reference) => reference.mediaKind === rule.mediaKind).reduce((total, reference) => total + (reference.durationMs ?? 0), 0) / 1e3 > rule.combinedDurationSeconds.max) return `${capability_.label} accepts at most ${rule.combinedDurationSeconds.max}s of ${rule.mediaKind} references combined`;
+		}
+	}
+	if (capability_.mode === "text" && references.length > 0) return "WAN 3.0 text-to-video does not accept references";
+	return null;
+}
+//#endregion
 //#region src/shared/videoGenerationOptions.ts
 var VIDEO_GENERATION_ASPECT_RATIOS = ["16:9", "9:16"];
 var VIDEO_GENERATION_RESOLUTIONS = [
@@ -6380,13 +6564,17 @@ var VIDEO_MODEL_SELECTIONS = [
 	"veo-3.1",
 	"omni-flash",
 	"kling",
-	"fal-seedance"
+	"fal-seedance",
+	...WAN_3_SELECTIONS
 ];
 var VIDEO_MODEL_LABELS = {
 	"veo-3.1": "Veo 3.1",
 	"omni-flash": "Omni Flash",
 	kling: "Kling 3.0",
-	"fal-seedance": "fal.ai Seedance v1 (text-to-video)"
+	"fal-seedance": "fal.ai Seedance v1 (text-to-video)",
+	"wan-3-text": "WAN 3.0 Text",
+	"wan-3-frame": "WAN 3.0 Frames",
+	"wan-3-reference": "WAN 3.0 References"
 };
 var VIDEO_MODEL_SUPPORTED_OPERATIONS = {
 	"veo-3.1": [
@@ -6404,7 +6592,10 @@ var VIDEO_MODEL_SUPPORTED_OPERATIONS = {
 		"refine",
 		"derive"
 	],
-	"fal-seedance": ["generate"]
+	"fal-seedance": ["generate"],
+	"wan-3-text": ["generate"],
+	"wan-3-frame": ["derive"],
+	"wan-3-reference": ["derive", "refine"]
 };
 var KLING_VIDEO_MODEL = "kling-v3";
 var FAL_VIDEO_MODEL = "fal-ai/bytedance/seedance/v1/pro/text-to-video";
@@ -6462,6 +6653,8 @@ function getVideoGenerationModelForSelection(selection = DEFAULT_VIDEO_MODEL_SEL
 	if (selection === "omni-flash") return GEMINI_OMNI_FLASH_VIDEO_MODEL;
 	if (selection === "kling") return KLING_VIDEO_MODEL;
 	if (selection === "fal-seedance") return FAL_VIDEO_MODEL;
+	const wan3 = getWan3CapabilityBySelection(selection);
+	if (wan3) return wan3.endpointId;
 	return getVideoGenerationModelForTier(tier);
 }
 function resolveVideoGenerationModel(model, tier = DEFAULT_VIDEO_GENERATION_TIER) {
@@ -6477,6 +6670,8 @@ function isVideoGenerationResolutionSupportedForTier(resolution, tier = DEFAULT_
 	return VIDEO_GENERATION_RESOLUTIONS_BY_TIER[tier].includes(resolution);
 }
 function getVideoGenerationResolutionsForModel(model) {
+	const wan3 = getWan3CapabilityByEndpoint(model);
+	if (wan3) return [...wan3.resolutions];
 	if (isKlingVideoGenerationModel(model) || isFalSeedanceV1VideoGenerationModel(model)) return KLING_VIDEO_GENERATION_RESOLUTIONS;
 	const seedanceCapability = getSeedance2CapabilityByEndpoint(model);
 	if (seedanceCapability) return [...seedanceCapability.resolutions];
@@ -6493,6 +6688,8 @@ function getVideoModelSelectionForModel(model) {
 	if (model === "gemini-omni-flash-preview") return "omni-flash";
 	if (model === "kling-v3") return "kling";
 	if (model === "fal-ai/bytedance/seedance/v1/pro/text-to-video") return "fal-seedance";
+	const wan3 = getWan3CapabilityByEndpoint(model);
+	if (wan3) return wan3.selection;
 	return getVideoGenerationTierForModel(model) ? "veo-3.1" : void 0;
 }
 function isVeoVideoGenerationModel(model) {
@@ -6505,7 +6702,7 @@ function isKlingVideoGenerationModel(model) {
 	return model === KLING_VIDEO_MODEL;
 }
 function isFalVideoGenerationModel(model) {
-	return isFalSeedanceV1VideoGenerationModel(model) || isFalSeedance2VideoGenerationModel(model);
+	return isFalSeedanceV1VideoGenerationModel(model) || isFalSeedance2VideoGenerationModel(model) || isFalWan3VideoGenerationModel(model);
 }
 function isFalSeedanceV1VideoGenerationModel(model) {
 	return model === FAL_VIDEO_MODEL;
@@ -6513,10 +6710,15 @@ function isFalSeedanceV1VideoGenerationModel(model) {
 function isFalSeedance2VideoGenerationModel(model) {
 	return typeof model === "string" && getSeedance2CapabilityByEndpoint(model) !== void 0;
 }
+function isFalWan3VideoGenerationModel(model) {
+	return getWan3CapabilityByEndpoint(model) !== void 0;
+}
 function isVideoGenerationModel(model) {
 	return isVeoVideoGenerationModel(model) || isGeminiOmniVideoGenerationModel(model) || isKlingVideoGenerationModel(model) || isFalVideoGenerationModel(model);
 }
 function getVideoGenerationMaxReferenceImages(model) {
+	const wan3 = getWan3CapabilityByEndpoint(model);
+	if (wan3) return wan3.references.find((reference) => reference.mediaKind === "image")?.maxCount ?? 0;
 	if (isFalSeedanceV1VideoGenerationModel(model)) return 0;
 	if (isFalSeedance2VideoGenerationModel(model)) return getSeedance2CapabilityByEndpoint(model)?.references.find((reference) => reference.mediaKind === "image")?.maxCount ?? 0;
 	if (isKlingVideoGenerationModel(model)) return 1;
@@ -6529,7 +6731,7 @@ function isVideoOperationSupportedByModel(model, operation) {
 	return selection !== void 0 && VIDEO_MODEL_SUPPORTED_OPERATIONS[selection].includes(operation);
 }
 function doesVideoGenerationModelSupportAudioToggle(model) {
-	return isKlingVideoGenerationModel(model) || isVeoVideoGenerationModel(model) && VIDEO_GENERATION_AUDIO_TOGGLE_MODELS.includes(model);
+	return isFalWan3VideoGenerationModel(model) || isKlingVideoGenerationModel(model) || isVeoVideoGenerationModel(model) && VIDEO_GENERATION_AUDIO_TOGGLE_MODELS.includes(model);
 }
 //#endregion
 //#region src/shared/generationRouting.ts
@@ -7047,6 +7249,9 @@ var CLI_GENERATION_MEDIA_OPTIONS = {
 		"mode",
 		"guidance-scale",
 		"audio-guidance-scale",
+		"seed",
+		"prompt-expansion",
+		"thinking",
 		"provider"
 	]
 };
@@ -7103,23 +7308,28 @@ async function executeGenerate(parsed, ctx, client, deps, mediaKind, followOptio
 	const musicProvider = parseMusicProviderOption(parsed, mediaKind, assetType);
 	const effectiveVideoModel = videoOptions.model ?? getVideoGenerationModelForSelection();
 	const seedanceCapability = mediaKind === "video" ? getSeedance2CapabilityByEndpoint(effectiveVideoModel) : void 0;
+	const wan3Capability = mediaKind === "video" ? getWan3CapabilityByEndpoint(effectiveVideoModel) : void 0;
 	const videoFrameRefs = parseVideoFrameReferenceOptions(parsed, "generate", mediaKind).refs;
 	const seedanceRefs = parseSeedanceReferenceOptions(parsed);
+	const seedanceReferenceCount = seedanceRefs.imageRefs.length + seedanceRefs.videoRefs.length + seedanceRefs.audioRefs.length;
 	const plainRefs = parseOptionalRefs(parsed, "refs");
 	validateSeedanceReferenceOptions(seedanceRefs, effectiveVideoModel, "generate");
 	if (seedanceCapability?.mode === "frame" && videoFrameRefs.length === 0) throw new Error(`--model ${seedanceCapability.selection} requires --first-frame`);
 	if (seedanceCapability?.mode === "text" && videoFrameRefs.length > 0) throw new Error(`--model ${seedanceCapability.selection} does not accept references`);
 	if (seedanceCapability?.mode === "reference" && videoFrameRefs.length > 0) throw new Error(`--model ${seedanceCapability.selection} uses --image-refs, --video-refs, and --audio-refs`);
 	if (seedanceCapability && plainRefs.length > 0) throw new Error(`--model ${seedanceCapability.selection} uses --image-refs, --video-refs, and --audio-refs instead of --refs`);
+	if (wan3Capability?.mode === "frame" && videoFrameRefs.length === 0) throw new Error(`--model ${wan3Capability.selection} requires --first-frame`);
+	if (wan3Capability?.mode === "text" && (videoFrameRefs.length > 0 || seedanceReferenceCount > 0)) throw new Error(`--model ${wan3Capability.selection} does not accept references`);
+	if (wan3Capability?.mode === "reference" && videoFrameRefs.length > 0) throw new Error(`--model ${wan3Capability.selection} uses --image-refs, --video-refs, and --audio-refs`);
+	if (wan3Capability && plainRefs.length > 0) throw new Error(`--model ${wan3Capability.selection} uses typed frame or grouped reference options instead of --refs`);
 	if (mediaKind === "video" && isKlingVideoGenerationModel(effectiveVideoModel) && videoFrameRefs.length + plainRefs.length > 1) throw new Error("--model kling supports at most one image reference");
-	if (mediaKind === "video" && isFalVideoGenerationModel(effectiveVideoModel) && !isFalSeedance2VideoGenerationModel(effectiveVideoModel) && (videoFrameRefs.length > 0 || plainRefs.length > 0)) throw new Error(`--model ${getVideoModelSelectionForModel(effectiveVideoModel) ?? effectiveVideoModel} does not support image references`);
-	const seedanceReferenceCount = seedanceRefs.imageRefs.length + seedanceRefs.videoRefs.length + seedanceRefs.audioRefs.length;
+	if (mediaKind === "video" && isFalVideoGenerationModel(effectiveVideoModel) && !isFalSeedance2VideoGenerationModel(effectiveVideoModel) && !isFalWan3VideoGenerationModel(effectiveVideoModel) && (videoFrameRefs.length > 0 || plainRefs.length > 0)) throw new Error(`--model ${getVideoModelSelectionForModel(effectiveVideoModel) ?? effectiveVideoModel} does not support image references`);
 	const state = videoFrameRefs.length > 0 || seedanceReferenceCount > 0 || plainRefs.length > 0 ? await requestSpaceState(client) : void 0;
 	const referenceDeps = {
 		...deps,
 		waitForReferenceVariant: (variant) => waitForReferenceVariant(client, variant)
 	};
-	const resolvedSeedanceRefs = state && seedanceReferenceCount > 0 ? isAvatarModelId(effectiveVideoModel) ? await resolveAvatarReferenceIds(seedanceRefs, ctx, referenceDeps, state) : await resolveSeedanceReferenceIds(seedanceRefs, seedanceCapability, ctx, referenceDeps, state) : {
+	const resolvedSeedanceRefs = state && seedanceReferenceCount > 0 ? isAvatarModelId(effectiveVideoModel) ? await resolveAvatarReferenceIds(seedanceRefs, ctx, referenceDeps, state) : wan3Capability ? await resolveWan3ReferenceIds(seedanceRefs, ctx, referenceDeps, state) : await resolveSeedanceReferenceIds(seedanceRefs, seedanceCapability, ctx, referenceDeps, state) : {
 		refs: [],
 		ids: []
 	};
@@ -7298,11 +7508,26 @@ function validateSeedanceReferenceOptions(options, model, command) {
 		return;
 	}
 	const capability = getSeedance2CapabilityByEndpoint(model);
+	const wan3Capability = getWan3CapabilityByEndpoint(model);
 	const total = options.imageRefs.length + options.videoRefs.length + options.audioRefs.length;
-	if (!capability) {
-		if (total > 0) throw new Error("--image-refs, --video-refs, and --audio-refs require a Seedance 2 reference model");
+	if (!capability && !wan3Capability) {
+		if (total > 0) throw new Error("--image-refs, --video-refs, and --audio-refs require a Seedance 2 or WAN 3.0 reference model");
 		return;
 	}
+	if (wan3Capability) {
+		if (wan3Capability.mode !== "reference") {
+			if (total > 0) throw new Error(`--model ${wan3Capability.selection} does not accept grouped references`);
+			return;
+		}
+		const error = getWan3ReferenceError(wan3Capability, [
+			...options.imageRefs.map(() => ({ mediaKind: "image" })),
+			...options.videoRefs.map(() => ({ mediaKind: "video" })),
+			...options.audioRefs.map(() => ({ mediaKind: "audio" }))
+		]);
+		if (error) throw new Error(error);
+		return;
+	}
+	if (!capability) return;
 	if (capability.mode !== "reference") {
 		if (total > 0) throw new Error(`--model ${capability.selection} does not accept grouped references`);
 		return;
@@ -7345,6 +7570,26 @@ async function resolveSeedanceReferenceIds(options, capability, ctx, deps, state
 		]
 	};
 }
+async function resolveWan3ReferenceIds(options, ctx, deps, state) {
+	const resolve = (refs, kind) => resolveReferenceVariantIds(refs, ctx, deps, state.variants, "video", state.assets, kind);
+	const [imageIds, videoIds, audioIds] = await Promise.all([
+		resolve(options.imageRefs, "image"),
+		resolve(options.videoRefs, "video"),
+		resolve(options.audioRefs, "audio")
+	]);
+	return {
+		refs: [
+			...options.imageRefs,
+			...options.videoRefs,
+			...options.audioRefs
+		],
+		ids: [
+			...imageIds,
+			...videoIds,
+			...audioIds
+		]
+	};
+}
 async function preflightLocalSeedanceReferences(options, deps, state, capability = SEEDANCE_2_CAPABILITIES["seedance-2-reference"]) {
 	if (!deps.inspectLocalReference) return;
 	const inspections = {
@@ -7690,9 +7935,12 @@ function parseVideoGenerationOptions(parsed, mediaKind) {
 	const modeValue = readOptionalOption(parsed, "mode");
 	const guidanceScaleValue = readOptionalOption(parsed, "guidance-scale");
 	const audioGuidanceScaleValue = readOptionalOption(parsed, "audio-guidance-scale");
+	const seedValue = mediaKind === "video" ? readOptionalOption(parsed, "seed") : void 0;
+	const promptExpansionValue = mediaKind === "video" ? parsed.options["prompt-expansion"] : void 0;
+	const thinkingValue = mediaKind === "video" ? parsed.options.thinking : void 0;
 	const modelValue = mediaKind === "video" ? readOptionalOption(parsed, "model") : void 0;
 	const aspectValue = mediaKind === "video" ? readOptionalOption(parsed, "aspect") : void 0;
-	if (mediaKind !== "video" && resolutionValue === void 0 && durationValue === void 0 && tierValue === void 0 && aspectValue === void 0 && bitrateValue === void 0 && modeValue === void 0 && guidanceScaleValue === void 0 && audioGuidanceScaleValue === void 0) return {};
+	if (mediaKind !== "video" && resolutionValue === void 0 && durationValue === void 0 && tierValue === void 0 && aspectValue === void 0 && bitrateValue === void 0 && modeValue === void 0 && guidanceScaleValue === void 0 && audioGuidanceScaleValue === void 0 && seedValue === void 0 && promptExpansionValue === void 0 && thinkingValue === void 0) return {};
 	if (mediaKind !== "video") throw new Error("Video model controls are only supported for video generation");
 	const videoResolution = resolutionValue === void 0 ? void 0 : normalizeVideoGenerationResolution(resolutionValue);
 	if (resolutionValue !== void 0 && !videoResolution) throw new Error("--resolution must be 480p, 720p, 1080p, or 4k");
@@ -7701,7 +7949,7 @@ function parseVideoGenerationOptions(parsed, mediaKind) {
 	const model = parseVideoModelOption(modelValue, videoTier, parsed);
 	if (isAvatarModelId(model)) {
 		if (parsed.options.refs !== void 0 || parsed.options["first-frame"] !== void 0 || parsed.options["last-frame"] !== void 0) throw new Error("Avatar models accept only --image-refs and --audio-refs");
-		if (aspectValue !== void 0 || durationValue !== void 0 || tierValue !== void 0 || bitrateValue !== void 0 || parsed.options.audio !== void 0 || parsed.options["no-audio"] !== void 0) throw new Error("Avatar models do not accept --aspect, --duration, --tier, --bitrate, --audio, or --no-audio");
+		if (aspectValue !== void 0 || durationValue !== void 0 || tierValue !== void 0 || bitrateValue !== void 0 || parsed.options.audio !== void 0 || parsed.options["no-audio"] !== void 0 || seedValue !== void 0 || promptExpansionValue !== void 0 || thinkingValue !== void 0) throw new Error("Avatar models do not accept --aspect, --duration, --tier, --bitrate, --audio, or --no-audio");
 		if (model === "kling-avatar-v2") {
 			if (parsed.options.provider !== void 0) throw new Error("Kling Avatar V2 does not accept --provider");
 			if (resolutionValue !== void 0 || guidanceScaleValue !== void 0 || audioGuidanceScaleValue !== void 0) throw new Error("Aurora controls are not supported by --model kling-avatar-v2");
@@ -7731,26 +7979,39 @@ function parseVideoGenerationOptions(parsed, mediaKind) {
 	if (modeValue !== void 0) throw new Error("--mode is only supported by --model kling-avatar-v2");
 	if (guidanceScaleValue !== void 0 || audioGuidanceScaleValue !== void 0) throw new Error("--guidance-scale and --audio-guidance-scale require --model creatify-aurora");
 	const seedanceCapability = getSeedance2CapabilityByEndpoint(model);
+	const wan3Capability = getWan3CapabilityByEndpoint(model);
 	const seedanceDuration = durationValue === void 0 ? void 0 : seedanceCapability ? normalizeSeedanceDuration(durationValue, model) : void 0;
+	const wan3Duration = durationValue === void 0 ? void 0 : wan3Capability ? durationValue === "auto" ? "auto" : Number(durationValue) : void 0;
 	const videoDurationSeconds = durationValue === void 0 || durationValue === "auto" ? void 0 : normalizeVideoGenerationDurationSeconds(durationValue);
 	if (seedanceCapability && durationValue !== void 0 && seedanceDuration === void 0) {
 		const numericDurations = seedanceCapability.durations.filter((value) => typeof value === "number");
 		throw new Error(`--duration must be auto or an integer from ${numericDurations[0]} to ${numericDurations.at(-1)}`);
 	}
-	if (!seedanceCapability && durationValue === "auto") throw new Error("--duration auto is only supported by Seedance 2");
+	if (wan3Capability && durationValue !== void 0 && !isWan3Duration(wan3Duration)) throw new Error("--duration must be auto or an integer from 2 to 30 for WAN 3.0");
+	if (!seedanceCapability && !wan3Capability && durationValue === "auto") throw new Error("--duration auto is only supported by Seedance 2 and WAN 3.0");
 	if (isGeminiOmniVideoGenerationModel(model) && (resolutionValue !== void 0 || durationValue !== void 0 || tierValue !== void 0)) throw new Error("--resolution, --duration, and --tier are only supported with --model veo-3.1");
 	if ((isKlingVideoGenerationModel(model) || isFalVideoGenerationModel(model)) && tierValue !== void 0) throw new Error("--tier is only supported with --model veo-3.1");
 	if (videoResolution && videoTier && !isVideoGenerationResolutionSupportedForTier(videoResolution, videoTier)) throw new Error("--resolution 4k is not supported with --tier lite");
 	if (videoResolution && (isKlingVideoGenerationModel(model) || isFalVideoGenerationModel(model)) && !isVideoGenerationResolutionSupportedForModel(model, videoResolution)) throw new Error(`--model ${getVideoModelSelectionForModel(model) ?? model} does not support --resolution ${videoResolution}`);
-	if (!seedanceCapability && videoDurationSeconds && ![
+	if (!seedanceCapability && !wan3Capability && videoDurationSeconds && ![
 		4,
 		6,
 		8
 	].includes(videoDurationSeconds)) throw new Error("--duration must be 4, 6, or 8 for this model");
-	const aspectRatio = aspectValue === void 0 ? void 0 : seedanceCapability ? seedanceCapability.aspectRatios.includes(aspectValue) ? aspectValue : void 0 : normalizeVideoGenerationAspectRatio(aspectValue);
-	if (aspectValue !== void 0 && !aspectRatio) throw new Error(seedanceCapability ? `--aspect must be one of: ${seedanceCapability.aspectRatios.join(", ")}` : "--aspect must be 16:9 or 9:16");
+	const aspectRatio = aspectValue === void 0 ? void 0 : seedanceCapability ? seedanceCapability.aspectRatios.includes(aspectValue) ? aspectValue : void 0 : wan3Capability ? wan3Capability.aspectRatios.includes(aspectValue) ? aspectValue : void 0 : normalizeVideoGenerationAspectRatio(aspectValue);
+	if (aspectValue !== void 0 && !aspectRatio) throw new Error(seedanceCapability ? `--aspect must be one of: ${seedanceCapability.aspectRatios.join(", ")}` : wan3Capability ? `--aspect must be one of: ${wan3Capability.aspectRatios.join(", ")}` : "--aspect must be 16:9 or 9:16");
 	const seedanceBitrateMode = bitrateValue;
 	if (bitrateValue !== void 0 && (!seedanceCapability || !seedanceCapability.bitrateModes.includes(seedanceBitrateMode))) throw new Error(seedanceCapability ? `${seedanceCapability.familyLabel} does not accept --bitrate` : "--bitrate standard|high is only supported by Seedance 2");
+	const wan3Seed = seedValue === void 0 ? void 0 : Number(seedValue);
+	if (seedValue !== void 0 && (!wan3Capability || !Number.isSafeInteger(wan3Seed) || wan3Seed < 0 || wan3Seed > 2147483647)) throw new Error(wan3Capability ? "--seed must be an integer from 0 to 2147483647" : "--seed is only supported by WAN 3.0 for video generation");
+	const parseWanBoolean = (value, option) => {
+		if (value === void 0) return void 0;
+		if (!wan3Capability) throw new Error(`${option} is only supported by WAN 3.0`);
+		if (value !== "true" && value !== "false") throw new Error(`${option} must be true or false`);
+		return value === "true";
+	};
+	const wan3PromptExpansion = parseWanBoolean(promptExpansionValue, "--prompt-expansion");
+	const wan3Thinking = parseWanBoolean(thinkingValue, "--thinking");
 	return {
 		model,
 		...aspectRatio ? { aspectRatio } : {},
@@ -7758,7 +8019,11 @@ function parseVideoGenerationOptions(parsed, mediaKind) {
 		...videoDurationSeconds ? { videoDurationSeconds } : {},
 		...videoTier ? { videoTier } : {},
 		...seedanceCapability && seedanceDuration !== void 0 ? { seedanceDuration } : {},
-		...seedanceBitrateMode ? { seedanceBitrateMode } : {}
+		...seedanceBitrateMode ? { seedanceBitrateMode } : {},
+		...wan3Capability && wan3Duration !== void 0 ? { wan3Duration } : {},
+		...wan3Seed !== void 0 ? { wan3Seed } : {},
+		...wan3PromptExpansion !== void 0 ? { wan3PromptExpansion } : {},
+		...wan3Thinking !== void 0 ? { wan3Thinking } : {}
 	};
 }
 function parseGenerationProviderOption(parsed, mediaKind, videoModel) {
@@ -7782,12 +8047,13 @@ function validateVideoFrameReferenceOptions(command, parsed, mediaKind) {
 	if (mediaKind !== "video") throw new Error("--first-frame and --last-frame are only supported for video generation");
 	if (lastFrame && !firstFrame) throw new Error("--last-frame requires --first-frame");
 	if (parsed.options.refs) throw new Error("--first-frame and --last-frame cannot be combined with --refs");
-	if (parsed.options["image-refs"] || parsed.options["video-refs"] || parsed.options["audio-refs"]) throw new Error("Seedance 2 cannot combine authoritative --first-frame/--last-frame inputs with extended --image-refs/--video-refs/--audio-refs; choose frame mode or reference mode");
+	if (parsed.options["image-refs"] || parsed.options["video-refs"] || parsed.options["audio-refs"]) throw new Error("Frame mode cannot combine --first-frame/--last-frame with grouped --image-refs/--video-refs/--audio-refs");
 	const tierValue = readOptionalOption(parsed, "tier");
 	const videoTier = tierValue === void 0 ? void 0 : normalizeVideoGenerationTier(tierValue);
 	const model = parseVideoModelOption(readOptionalOption(parsed, "model"), videoTier, parsed);
+	const wan3Capability = getWan3CapabilityByEndpoint(model);
 	if (isFalSeedance2VideoGenerationModel(model) && !firstFrame) throw new Error("Seedance frame mode requires --first-frame");
-	if (!isFalSeedance2VideoGenerationModel(model) && (isFalVideoGenerationModel(model) || isGeminiOmniVideoGenerationModel(model) || isKlingVideoGenerationModel(model) && Boolean(lastFrame))) throw new Error("--first-frame and --last-frame require a frame-capable video model");
+	if (!(isFalSeedance2VideoGenerationModel(model) || wan3Capability?.mode === "frame") && (isFalVideoGenerationModel(model) || isGeminiOmniVideoGenerationModel(model) || isKlingVideoGenerationModel(model) && Boolean(lastFrame))) throw new Error("--first-frame and --last-frame require a frame-capable video model");
 }
 function readVideoFrameOption(parsed, name) {
 	const value = parsed.options[name];
@@ -7854,9 +8120,14 @@ function parseVideoModelOption(value, tier, parsed) {
 		const hasReferences = Boolean(parsed?.options.refs || parsed?.options["image-refs"] || parsed?.options["video-refs"] || parsed?.options["audio-refs"]);
 		return SEEDANCE_2_CAPABILITIES[`${normalized}-${hasFrames ? "frame" : hasReferences ? "reference" : "text"}`].endpointId;
 	}
+	if (normalized === "wan-3") {
+		const hasFrames = Boolean(parsed?.options["first-frame"] || parsed?.options["last-frame"]);
+		const hasReferences = Boolean(parsed?.options["image-refs"] || parsed?.options["video-refs"] || parsed?.options["audio-refs"]);
+		return getWan3CapabilityBySelection(`wan-3-${hasFrames ? "frame" : hasReferences ? "reference" : "text"}`).endpointId;
+	}
 	const selection = normalizeVideoModelSelection(normalized === "seedance-1" ? "fal-seedance" : normalized);
 	if (selection) return getVideoGenerationModelForSelection(selection, effectiveTier);
-	throw new Error("--model must be veo-3.1, omni-flash, kling, seedance-1, seedance-2, seedance-2-fast, seedance-2.5, kling-avatar-v2, or creatify-aurora");
+	throw new Error("--model must be veo-3.1, omni-flash, kling, seedance-1, seedance-2, seedance-2-fast, seedance-2.5, wan-3, kling-avatar-v2, or creatify-aurora");
 }
 function parseImageModelOption(value) {
 	if (!value) return void 0;
@@ -8038,6 +8309,8 @@ function parseWaitSeconds$1(value) {
 function buildRegenerationBody(mediaKind, parsed, referenceVariantIds, audioMode, sourceRecipe) {
 	const prompt = parsed.positionals[1] ?? parsed.options.prompt;
 	const model = parsed.options.model;
+	const storedRecipe = parseStoredRecipe$2(sourceRecipe);
+	const isWan3 = mediaKind === "video" && (model !== void 0 ? model === "wan-3" : Boolean(getWan3CapabilityByEndpoint(storedRecipe?.model)));
 	const audioModel = mediaKind === "audio" && (audioMode === "speech" || audioMode === "dialogue") ? normalizeElevenLabsSpeechModelId(model) : void 0;
 	const params = {};
 	let duration;
@@ -8057,11 +8330,23 @@ function buildRegenerationBody(mediaKind, parsed, referenceVariantIds, audioMode
 	} else if (mediaKind === "video") {
 		if (parsed.options.aspect) params.aspectRatio = parsed.options.aspect;
 		if (parsed.options.resolution) params.videoResolution = parsed.options.resolution;
-		if (parsed.options.duration) duration = parseInteger(parsed.options.duration, "--duration");
+		if (parsed.options.duration) if (isWan3) {
+			const wan3Duration = parsed.options.duration === "auto" ? "auto" : parseInteger(parsed.options.duration, "--duration");
+			if (!isWan3Duration(wan3Duration)) throw new Error("--duration must be auto or an integer from 2 to 30 for WAN 3.0");
+			params.wan3Duration = wan3Duration;
+		} else duration = parseInteger(parsed.options.duration, "--duration");
 		if (parsed.options.tier) params.videoTier = parsed.options.tier;
 		if (parsed.options.bitrate) params.seedanceBitrateMode = parsed.options.bitrate;
 		if (parsed.options.audio === "true") params.generateAudio = true;
 		if (parsed.options["no-audio"] === "true") params.generateAudio = false;
+		if ((parsed.options.seed !== void 0 || parsed.options["prompt-expansion"] !== void 0 || parsed.options.thinking !== void 0) && !isWan3) throw new Error("--seed, --prompt-expansion, and --thinking require --model wan-3 or a stored WAN 3.0 recipe");
+		if (parsed.options.seed !== void 0) {
+			const seed = parseInteger(parsed.options.seed, "--seed");
+			if (seed < 0 || seed > 2147483647) throw new Error("--seed must be an integer from 0 to 2147483647 for WAN 3.0");
+			params.wan3Seed = seed;
+		}
+		if (parsed.options["prompt-expansion"] !== void 0) params.wan3PromptExpansion = parseBooleanOption(parsed.options["prompt-expansion"], "--prompt-expansion");
+		if (parsed.options.thinking !== void 0) params.wan3Thinking = parseBooleanOption(parsed.options.thinking, "--thinking");
 	} else {
 		if (!audioMode) throw new Error("Audio regeneration mode is required");
 		validateAudioModel(audioMode, audioModel ?? model);
@@ -8084,6 +8369,7 @@ function buildRegenerationBody(mediaKind, parsed, referenceVariantIds, audioMode
 }
 function resolveVideoRegenerationModel(model, parsed, sourceRecipe) {
 	const tier = parsed.options.tier ?? "generate";
+	if (model === "wan-3") return getWan3CapabilityBySelection("wan-3-reference").endpointId;
 	if (model === "seedance-2" || model === "seedance-2-fast" || model === "seedance-2.5") {
 		const referencesWereOverridden = [
 			"image-refs",
@@ -8103,7 +8389,7 @@ function resolveVideoRegenerationModel(model, parsed, sourceRecipe) {
 		}, referencesWereOverridden).endpointId;
 	}
 	const selection = normalizeVideoModelSelection(model === "seedance-1" ? "fal-seedance" : model);
-	if (!selection) throw new Error("Unsupported video model. Expected veo-3.1, omni-flash, kling, seedance-1, seedance-2, seedance-2-fast, or seedance-2.5");
+	if (!selection) throw new Error("Unsupported video model. Expected veo-3.1, omni-flash, kling, seedance-1, seedance-2, seedance-2-fast, seedance-2.5, or wan-3");
 	return getVideoGenerationModelForSelection(selection, tier);
 }
 function parseStoredRecipe$2(recipe) {
@@ -8147,6 +8433,11 @@ function parseInteger(value, option) {
 	if (!Number.isInteger(parsed)) throw new Error(`${option} must be an integer`);
 	return parsed;
 }
+function parseBooleanOption(value, option) {
+	if (value === "true") return true;
+	if (value === "false") return false;
+	throw new Error(`${option} must be true or false`);
+}
 //#endregion
 //#region src/cli/commands/audio.ts
 var defaultDeps$10 = {
@@ -9331,6 +9622,124 @@ function seedanceVideoGenerator(capability) {
 		]
 	};
 }
+function wan3VideoGenerator(capability) {
+	const fixedGenerator = fixedInput("generator_id", capability.generatorId, "Selects this exact WAN 3.0 mode.");
+	const parameters = [
+		input("aspect_ratio", "string", false, "Output aspect ratio; adaptive lets WAN choose.", {
+			allowedValues: capability.aspectRatios,
+			defaultValue: capability.defaultAspectRatio
+		}),
+		input("resolution", "string", false, "Output resolution.", {
+			allowedValues: capability.resolutions,
+			defaultValue: capability.defaultResolution
+		}),
+		input("duration", "string_or_integer", false, "Output duration from 2 to 30 seconds, or auto.", {
+			allowedValues: capability.durations,
+			defaultValue: capability.defaultDuration
+		}),
+		input("generate_audio", "boolean", false, "Generate synchronized native audio.", {
+			allowedValues: [true, false],
+			defaultValue: true
+		}),
+		input("seed", "integer", false, "Optional deterministic provider seed.", {
+			minimum: 0,
+			maximum: 2147483647
+		}),
+		input("prompt_expansion", "boolean", false, "Allow WAN to expand the prompt for quality.", {
+			allowedValues: [true, false],
+			defaultValue: true
+		}),
+		input("thinking", "boolean", false, "Enable enhanced composition and motion reasoning.", {
+			allowedValues: [true, false],
+			defaultValue: false
+		})
+	];
+	const common = [
+		SPACE_INPUT,
+		fixedGenerator,
+		NAME_INPUT,
+		input("asset_type", "string", true, "Asset classification stored in the Space."),
+		PROMPT_INPUT,
+		...parameters
+	];
+	const max = (kind) => capability.references.find((rule) => rule.mediaKind === kind)?.maxCount ?? 0;
+	const referenceInputs = [
+		input("image_reference_variant_refs", "string_array", false, `Ordered completed images addressed as Image 1 through Image ${max("image")}.`, {
+			minItems: 1,
+			maxItems: max("image")
+		}),
+		input("video_reference_variant_refs", "string_array", false, `Ordered completed videos addressed as Video 1 through Video ${max("video")}; 15 seconds combined maximum.`, {
+			minItems: 1,
+			maxItems: max("video")
+		}),
+		input("audio_reference_variant_refs", "string_array", false, `Ordered completed audio clips addressed as Audio 1 through Audio ${max("audio")}; 15 seconds combined maximum.`, {
+			minItems: 1,
+			maxItems: max("audio")
+		})
+	];
+	const modeInputs = capability.mode === "frame" ? [input("start_frame_variant_ref", "string", true, "Required authoritative first frame."), input("end_frame_variant_ref", "string", false, "Optional authoritative final frame.")] : capability.mode === "reference" ? referenceInputs : [];
+	const operations = [{
+		operation: capability.mode === "text" ? "generate" : "derive",
+		tool: "generate_video",
+		description: capability.mode === "text" ? "Create a native-audio video from text." : capability.mode === "frame" ? "Animate a start frame and optional end frame." : "Direct one video from up to 20 ordered image, video, and audio references.",
+		inputs: [...common, ...modeInputs]
+	}];
+	if (capability.mode === "reference") operations.push({
+		operation: "refine",
+		tool: "edit_video",
+		description: "Refine a completed video as Video 1 with up to 19 additional references.",
+		referenceLimits: {
+			maxAdditionalCount: 19,
+			implicitSourceCount: 1,
+			maxAdditionalByKind: {
+				image: 10,
+				video: 4,
+				audio: 5
+			}
+		},
+		inputs: [
+			SPACE_INPUT,
+			fixedGenerator,
+			input("asset_ref", "string", true, "Target video asset reference."),
+			input("source_variant_ref", "string", true, "Completed target video used as Video 1."),
+			PROMPT_INPUT,
+			...parameters,
+			...referenceInputs.map((item) => item.name === "video_reference_variant_refs" ? {
+				...item,
+				maxItems: 4,
+				description: "Up to 4 additional videos; the source video is Video 1."
+			} : item)
+		]
+	});
+	return {
+		id: capability.generatorId,
+		label: capability.label,
+		mediaKind: "video",
+		modelIds: [capability.endpointId],
+		defaultModelId: capability.endpointId,
+		operations,
+		referenceRules: {
+			mediaKind: capability.mode === "frame" ? "image" : null,
+			completedOnly: capability.mode !== "text",
+			maxCount: capability.maxReferenceFiles,
+			maxTotalCount: capability.maxReferenceFiles,
+			modalities: capability.references.map((rule) => ({
+				mediaKind: rule.mediaKind,
+				minCount: rule.minCount,
+				maxCount: rule.maxCount,
+				promptLabel: rule.promptLabel,
+				acceptedMimeTypes: rule.acceptedMimeTypes,
+				maxBytesPerFile: rule.maxBytesPerFile,
+				...rule.combinedDurationSeconds ? { combinedDurationSeconds: rule.combinedDurationSeconds } : {}
+			}))
+		},
+		notes: [
+			"WAN 3.0 generates native synchronized audio and supports 480p, 720p, and 1080p.",
+			"Safety checking remains enabled by MakeFX and is not a user-controlled option.",
+			capability.mode === "reference" ? "Reference mode accepts up to 10 images, 5 videos, and 5 audio clips (20 total)." : capability.mode === "frame" ? "Frame mode accepts exactly one start frame and one optional end frame." : "Text mode rejects all media references."
+		]
+	};
+}
 function avatarVideoGenerator(model) {
 	const capability = getAvatarModelCapabilities(model);
 	const parameters = capability.model === "kling-avatar-v2" ? [input("mode", "string", false, "Generation quality.", {
@@ -9451,8 +9860,9 @@ function getGeneratorCatalog(overrides = {}) {
 	const lyria = overrides.lyria ?? "lyria-3-clip-preview";
 	return [
 		...Object.values(IMAGE_MODEL_CAPABILITIES).map(imageGenerator),
-		...VIDEO_MODEL_SELECTIONS.map(videoGenerator),
+		...VIDEO_MODEL_SELECTIONS.filter((selection) => !getWan3CapabilityBySelection(selection)).map(videoGenerator),
 		...SEEDANCE_2_SELECTIONS.map((selection) => seedanceVideoGenerator(SEEDANCE_2_CAPABILITIES[selection])),
+		...WAN_3_CAPABILITIES.map(wan3VideoGenerator),
 		...AVATAR_MODEL_IDS.map(avatarVideoGenerator),
 		audioGenerator({
 			id: "audio/elevenlabs-speech",
@@ -9534,9 +9944,15 @@ var DRAFT_RECIPE_INPUT_KEYS = {
 	video_tier: "videoTier",
 	duration: "seedanceDuration",
 	bitrate_mode: "seedanceBitrateMode",
+	prompt_expansion: "wan3PromptExpansion",
+	thinking: "wan3Thinking",
 	provider: "provider"
 };
 function draftRecipeInputKey(inputName, generatorId) {
+	if (generatorId.startsWith("video/wan-3-")) {
+		if (inputName === "duration") return "wan3Duration";
+		if (inputName === "seed") return "wan3Seed";
+	}
 	if ((generatorId === "video/p-video-avatar" || generatorId === "video/creatify-aurora") && inputName === "resolution") return "avatarResolution";
 	if (generatorId === "video/kling-avatar-v2" && inputName === "mode") return "avatarMode";
 	if (generatorId === "video/creatify-aurora" && inputName === "guidance_scale") return "avatarGuidanceScale";
@@ -11970,17 +12386,21 @@ var HELP = {
   makefx image generate "prompt" --name <name> --type <type> -o <file> [--model pro|flash|flux|gpt-image-2|seedream-5-pro|seedream-5-lite] [--provider <provider>] [--refs <variant-ref-or-file,...>] [--aspect <ratio>] [--size 1K|2K|3K|4K] [--quality low|medium|high] [--seed <integer>] [--collection <id>] [--space <id>]
   makefx image regenerate <variant-ref> ["prompt"] [--model pro|flash|flux|gpt-image-2|seedream-5-pro|seedream-5-lite] [--provider <provider>] [--aspect <ratio>] [--size <size>] [--quality low|medium|high] [--seed <integer|random>] [--refs <refs>] [--no-activate] [--wait]`,
 	video: `Usage:
-  makefx video generate "prompt" --name <name> --type <type> -o <file> [--model veo-3.1|omni-flash|kling|seedance-1|seedance-2|seedance-2-fast|seedance-2.5|kling-avatar-v2|creatify-aurora] [--provider fal|elevenlabs|pika]
+  makefx video generate "prompt" --name <name> --type <type> -o <file> [--model veo-3.1|omni-flash|kling|seedance-1|seedance-2|seedance-2-fast|seedance-2.5|wan-3|kling-avatar-v2|creatify-aurora] [--provider fal|elevenlabs|pika]
       [--refs <variant-ref-or-file,...>] [--first-frame <ref>] [--last-frame <ref>] [--image-refs <refs>] [--video-refs <refs>] [--audio-refs <refs>]
       [--aspect <ratio>] [--resolution 480p|720p|1080p|4k] [--duration <seconds|auto>] [--tier generate|fast|lite] [--bitrate standard|high]
       [--audio | --no-audio] [--collection <id>] [--space <id>]
+  makefx video generate "prompt" --model wan-3 --name <name> --type <type> -o <file>
+      [--first-frame <ref> [--last-frame <ref>] | --image-refs <refs> --video-refs <refs> --audio-refs <refs>]
+      [--aspect adaptive|16:9|4:3|1:1|3:4|9:16] [--resolution 480p|720p|1080p] [--duration 2..30|auto]
+      [--audio | --no-audio] [--seed 0..2147483647] [--prompt-expansion true|false] [--thinking true|false]
   makefx video generate ["prompt"] --model kling-avatar-v2 --image-refs <portrait-ref> --audio-refs <audio-ref>
       --name <name> --type <type> -o <file> [--mode standard|pro] [--collection <id>] [--space <id>]
   makefx video generate ["prompt"] --model creatify-aurora --image-refs <portrait-ref> --audio-refs <audio-ref>
       --name <name> --type <type> -o <file> [--resolution 480p|720p] [--guidance-scale 0..5] [--audio-guidance-scale 0..5] [--collection <id>] [--space <id>]
-  makefx video regenerate <variant-ref> ["prompt"] [--model <model>] [--provider fal|elevenlabs|pika] [--aspect <ratio>] [--resolution <value>] [--duration <seconds>] [--tier <tier>] [--bitrate standard|high] [--audio | --no-audio] [--image-refs <refs>] [--video-refs <refs>] [--audio-refs <refs>] [--no-activate] [--wait]
+  makefx video regenerate <variant-ref> ["prompt"] [--model <model>] [--provider fal|elevenlabs|pika] [--aspect <ratio>] [--resolution <value>] [--duration <seconds|auto>] [--tier <tier>] [--bitrate standard|high] [--audio | --no-audio] [--seed 0..2147483647] [--prompt-expansion true|false] [--thinking true|false] [--image-refs <refs>] [--video-refs <refs>] [--audio-refs <refs>] [--no-activate] [--wait]
 
-Seedance mode is inferred: first/last frame selects frame mode, image/video/audio refs select reference mode, and no references selects text mode.`,
+Seedance and WAN mode are inferred: first/last frame selects frame mode, image/video/audio refs select reference mode, and no references selects text mode.`,
 	audio: `Usage:
   makefx audio voices [--json]
   makefx audio align <variant-ref> "transcript" [--space <id>] [--json]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "makefx",
-  "version": "1.6.6",
+  "version": "1.6.8",
   "description": "Command-line interface for AI-assisted game asset production with Make Effects.",
   "license": "MIT",
   "type": "module",