mulmocast 2.0.0 → 2.0.2

package/lib/actions/image_agents.js

@@ -76,7 +76,7 @@ export const imagePreprocessAgent = async (namedInputs) => {
             ...returnValue,
             // imagePath: isTypeMovie ? undefined : pluginPath,
             imagePath: isTypeMovie ? imagePath : pluginPath,
-            movieFile: isTypeMovie ? pluginPath : undefined,
+            movieFile: isTypeMovie ? pluginPath : returnValue.movieFile,
             imageFromMovie: isTypeMovie,
             referenceImageForMovie: pluginPath,
             markdown,

package/lib/agents/movie_genai_agent.js

@@ -2,7 +2,7 @@ import { readFileSync } from "fs";
 import { GraphAILogger, sleep } from "graphai";
 import { apiKeyMissingError, agentGenerationError, agentInvalidResponseError, imageAction, movieFileTarget, videoDurationTarget, hasCause, } from "../utils/error_cause.js";
 import { GoogleGenAI, PersonGeneration } from "@google/genai";
-import { getModelDuration } from "../utils/provider2agent.js";
+import { getModelDuration, provider2MovieAgent } from "../utils/provider2agent.js";
 export const getAspectRatio = (canvasSize) => {
     if (canvasSize.width > canvasSize.height) {
         return "16:9";
@@ -17,8 +17,7 @@ export const getAspectRatio = (canvasSize) => {
 export const movieGenAIAgent = async ({ namedInputs, params, config, }) => {
     const { prompt, imagePath, movieFile } = namedInputs;
     const aspectRatio = getAspectRatio(params.canvasSize);
-    const model = params.model ?? "veo-2.0-generate-001"; // "veo-3.0-generate-preview";
-    // const duration = params.duration ?? 8;
+    const model = params.model ?? provider2MovieAgent.google.defaultModel;
     const apiKey = config?.apiKey;
     if (!apiKey) {
         throw new Error("Google GenAI API key is required (GEMINI_API_KEY)", {
@@ -102,7 +101,6 @@ const movieGenAIAgentInfo = {
     category: ["movie"],
     author: "Receptron Team",
     repository: "https://github.com/receptron/mulmocast-cli/",
-    // source: "https://github.com/receptron/mulmocast-cli/blob/main/src/agents/image_google_agent.ts",
     license: "MIT",
     environmentVariables: [],
 };

package/lib/types/schema.d.ts

@@ -685,6 +685,12 @@ export declare const mulmoMovieParamsSchema: z.ZodObject<{
         [x: string]: string;
     }>>>;
     model: z.ZodOptional<z.ZodString>;
+    fillOption: z.ZodOptional<z.ZodObject<{
+        style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
+            aspectFit: "aspectFit";
+            aspectFill: "aspectFill";
+        }>>>;
+    }, z.core.$strip>>;
     transition: z.ZodOptional<z.ZodObject<{
         type: z.ZodEnum<{
             fade: "fade";
@@ -692,12 +698,6 @@ export declare const mulmoMovieParamsSchema: z.ZodObject<{
         }>;
         duration: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
     }, z.core.$strip>>;
-    fillOption: z.ZodOptional<z.ZodObject<{
-        style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
-            aspectFit: "aspectFit";
-            aspectFill: "aspectFill";
-        }>>>;
-    }, z.core.$strip>>;
 }, z.core.$strict>;
 export declare const mulmoPresentationStyleSchema: z.ZodObject<{
     $mulmocast: z.ZodObject<{
@@ -766,6 +766,12 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
             [x: string]: string;
         }>>>;
         model: z.ZodOptional<z.ZodString>;
+        fillOption: z.ZodOptional<z.ZodObject<{
+            style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
+                aspectFit: "aspectFit";
+                aspectFill: "aspectFill";
+            }>>>;
+        }, z.core.$strip>>;
         transition: z.ZodOptional<z.ZodObject<{
             type: z.ZodEnum<{
                 fade: "fade";
@@ -773,12 +779,6 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
             }>;
             duration: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
         }, z.core.$strip>>;
-        fillOption: z.ZodOptional<z.ZodObject<{
-            style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
-                aspectFit: "aspectFit";
-                aspectFill: "aspectFill";
-            }>>>;
-        }, z.core.$strip>>;
     }, z.core.$strict>>;
     soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
         provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -902,6 +902,12 @@ export declare const mulmoScriptSchema: z.ZodObject<{
             [x: string]: string;
         }>>>;
         model: z.ZodOptional<z.ZodString>;
+        fillOption: z.ZodOptional<z.ZodObject<{
+            style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
+                aspectFit: "aspectFit";
+                aspectFill: "aspectFill";
+            }>>>;
+        }, z.core.$strip>>;
         transition: z.ZodOptional<z.ZodObject<{
             type: z.ZodEnum<{
                 fade: "fade";
@@ -909,12 +915,6 @@ export declare const mulmoScriptSchema: z.ZodObject<{
             }>;
             duration: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
         }, z.core.$strip>>;
-        fillOption: z.ZodOptional<z.ZodObject<{
-            style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
-                aspectFit: "aspectFit";
-                aspectFill: "aspectFill";
-            }>>>;
-        }, z.core.$strip>>;
     }, z.core.$strict>>;
     soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
         provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -1314,6 +1314,12 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                 [x: string]: string;
             }>>>;
             model: z.ZodOptional<z.ZodString>;
+            fillOption: z.ZodOptional<z.ZodObject<{
+                style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
+                    aspectFit: "aspectFit";
+                    aspectFill: "aspectFill";
+                }>>>;
+            }, z.core.$strip>>;
             transition: z.ZodOptional<z.ZodObject<{
                 type: z.ZodEnum<{
                     fade: "fade";
@@ -1321,12 +1327,6 @@ export declare const mulmoStudioSchema: z.ZodObject<{
                 }>;
                 duration: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
             }, z.core.$strip>>;
-            fillOption: z.ZodOptional<z.ZodObject<{
-                style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
-                    aspectFit: "aspectFit";
-                    aspectFill: "aspectFill";
-                }>>>;
-            }, z.core.$strip>>;
         }, z.core.$strict>>;
         soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
             provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -1662,6 +1662,12 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
                 [x: string]: string;
             }>>>;
             model: z.ZodOptional<z.ZodString>;
+            fillOption: z.ZodOptional<z.ZodObject<{
+                style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
+                    aspectFit: "aspectFit";
+                    aspectFill: "aspectFill";
+                }>>>;
+            }, z.core.$strip>>;
             transition: z.ZodOptional<z.ZodObject<{
                 type: z.ZodEnum<{
                     fade: "fade";
@@ -1669,12 +1675,6 @@ export declare const mulmoPromptTemplateSchema: z.ZodObject<{
                 }>;
                 duration: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
             }, z.core.$strip>>;
-            fillOption: z.ZodOptional<z.ZodObject<{
-                style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
-                    aspectFit: "aspectFit";
-                    aspectFill: "aspectFill";
-                }>>>;
-            }, z.core.$strip>>;
         }, z.core.$strict>>;
         soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
             provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{
@@ -1792,6 +1792,12 @@ export declare const mulmoPromptTemplateFileSchema: z.ZodObject<{
                 [x: string]: string;
             }>>>;
             model: z.ZodOptional<z.ZodString>;
+            fillOption: z.ZodOptional<z.ZodObject<{
+                style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
+                    aspectFit: "aspectFit";
+                    aspectFill: "aspectFill";
+                }>>>;
+            }, z.core.$strip>>;
             transition: z.ZodOptional<z.ZodObject<{
                 type: z.ZodEnum<{
                     fade: "fade";
@@ -1799,12 +1805,6 @@ export declare const mulmoPromptTemplateFileSchema: z.ZodObject<{
                 }>;
                 duration: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
             }, z.core.$strip>>;
-            fillOption: z.ZodOptional<z.ZodObject<{
-                style: z.ZodDefault<z.ZodOptional<z.ZodEnum<{
-                    aspectFit: "aspectFit";
-                    aspectFill: "aspectFill";
-                }>>>;
-            }, z.core.$strip>>;
         }, z.core.$strict>>;
         soundEffectParams: z.ZodDefault<z.ZodOptional<z.ZodObject<{
             provider: z.ZodOptional<z.ZodDefault<z.ZodEnum<{

package/lib/types/schema.js

@@ -278,6 +278,11 @@ export const mulmoLipSyncParamsSchema = z.object({
     provider: z.string().optional(), // lip sync provider
     model: z.string().optional(), // default: provider specific
 });
+const movieParamsSchema = z.object({
+    provider: text2MovieProviderSchema.optional(),
+    model: z.string().optional(),
+    fillOption: mulmoFillOptionSchema.optional(), // for movie.ts
+});
 export const mulmoBeatSchema = z
     .object({
     speaker: speakerIdSchema.optional(),
@@ -289,11 +294,8 @@ export const mulmoBeatSchema = z
     duration: z.number().optional().describe("Duration of the beat. Used only when the text is empty"),
     imageParams: mulmoBeatImageParamsSchema.optional(), // beat specific parameters
     audioParams: beatAudioParamsSchema.optional(), // beat specific parameters
-    movieParams: z
-        .object({
-        provider: text2MovieProviderSchema.optional(),
-        model: z.string().optional(),
-        fillOption: mulmoFillOptionSchema.optional(), // for movie.ts
+    movieParams: movieParamsSchema
+        .extend({
         speed: z.number().optional().describe("Speed of the video. 1.0 is normal speed. 0.5 is half speed. 2.0 is double speed."), // for movie.ts
     })
         .optional(),
@@ -344,12 +346,9 @@ export const mulmoTransitionSchema = z.object({
     type: z.enum(["fade", "slideout_left"]),
     duration: z.number().min(0).max(2).optional().default(0.3), // transition duration in seconds
 });
-export const mulmoMovieParamsSchema = z
-    .object({
-    provider: text2MovieProviderSchema.optional(), // for agent
-    model: z.string().optional(), // default: provider specific. for agent
+export const mulmoMovieParamsSchema = movieParamsSchema
+    .extend({
     transition: mulmoTransitionSchema.optional(), // for movie.ts
-    fillOption: mulmoFillOptionSchema.optional(), // for movie.ts
 })
     .strict();
 export const mulmoPresentationStyleSchema = z.object({

package/lib/types/type.d.ts

@@ -1,5 +1,5 @@
 import { type CallbackFunction } from "graphai";
-import { langSchema, localizedTextSchema, mulmoBeatSchema, mulmoScriptSchema, mulmoStudioSchema, mulmoStudioBeatSchema, mulmoStoryboardSchema, mulmoStoryboardSceneSchema, mulmoStudioMultiLingualSchema, mulmoStudioMultiLingualArraySchema, mulmoStudioMultiLingualDataSchema, mulmoStudioMultiLingualFileSchema, speakerDictionarySchema, speakerSchema, mulmoSpeechParamsSchema, mulmoImageParamsSchema, mulmoImageParamsImagesValueSchema, mulmoImageParamsImagesSchema, mulmoFillOptionSchema, mulmoMovieParamsSchema, textSlideParamsSchema, speechOptionsSchema, speakerDataSchema, mulmoCanvasDimensionSchema, mulmoPromptTemplateSchema, mulmoPromptTemplateFileSchema, text2ImageProviderSchema, text2HtmlImageProviderSchema, text2MovieProviderSchema, text2SpeechProviderSchema, mulmoPresentationStyleSchema, multiLingualTextsSchema, mulmoImageAssetSchema, mulmoMermaidMediaSchema, mulmoTextSlideMediaSchema, mulmoMarkdownMediaSchema, mulmoImageMediaSchema, mulmoChartMediaSchema, mediaSourceSchema, mediaSourceMermaidSchema, mulmoSessionStateSchema, mulmoOpenAIImageModelSchema, mulmoGoogleImageModelSchema, mulmoGoogleMovieModelSchema, mulmoReplicateMovieModelSchema, mulmoImagePromptMediaSchema } from "./schema.js";
+import { langSchema, localizedTextSchema, mulmoBeatSchema, mulmoScriptSchema, mulmoStudioSchema, mulmoStudioBeatSchema, mulmoStoryboardSchema, mulmoStoryboardSceneSchema, mulmoStudioMultiLingualSchema, mulmoStudioMultiLingualArraySchema, mulmoStudioMultiLingualDataSchema, mulmoStudioMultiLingualFileSchema, speakerDictionarySchema, speakerSchema, mulmoSpeechParamsSchema, mulmoImageParamsSchema, mulmoImageParamsImagesValueSchema, mulmoImageParamsImagesSchema, mulmoFillOptionSchema, mulmoMovieParamsSchema, mulmoSoundEffectParamsSchema, mulmoLipSyncParamsSchema, textSlideParamsSchema, speechOptionsSchema, speakerDataSchema, mulmoCanvasDimensionSchema, mulmoPromptTemplateSchema, mulmoPromptTemplateFileSchema, text2ImageProviderSchema, text2HtmlImageProviderSchema, text2MovieProviderSchema, text2SpeechProviderSchema, mulmoPresentationStyleSchema, multiLingualTextsSchema, mulmoImageAssetSchema, mulmoMermaidMediaSchema, mulmoTextSlideMediaSchema, mulmoMarkdownMediaSchema, mulmoImageMediaSchema, mulmoChartMediaSchema, mediaSourceSchema, mediaSourceMermaidSchema, mulmoSessionStateSchema, mulmoOpenAIImageModelSchema, mulmoGoogleImageModelSchema, mulmoGoogleMovieModelSchema, mulmoReplicateMovieModelSchema, mulmoImagePromptMediaSchema } from "./schema.js";
 import { pdf_modes, pdf_sizes, storyToScriptGenerateMode } from "../utils/const.js";
 import type { LLM } from "../utils/provider2agent.js";
 import { z } from "zod";
@@ -37,6 +37,8 @@ export type MulmoStudioMultiLingualData = z.infer<typeof mulmoStudioMultiLingual
 export type MulmoStudioMultiLingualFile = z.infer<typeof mulmoStudioMultiLingualFileSchema>;
 export type MultiLingualTexts = z.infer<typeof multiLingualTextsSchema>;
 export type MulmoMovieParams = z.infer<typeof mulmoMovieParamsSchema>;
+export type MulmoSoundEffectParams = z.infer<typeof mulmoSoundEffectParamsSchema>;
+export type MulmoLipSyncParams = z.infer<typeof mulmoLipSyncParamsSchema>;
 export type MulmoOpenAIImageModel = z.infer<typeof mulmoOpenAIImageModelSchema>;
 export type MulmoGoogleImageModel = z.infer<typeof mulmoGoogleImageModelSchema>;
 export type MulmoGoogleMovieModel = z.infer<typeof mulmoGoogleMovieModelSchema>;

package/lib/utils/agent_error_handler.d.ts

@@ -0,0 +1,24 @@
+/**
+ * Wraps agent execution with standardized error handling
+ *
+ * @param agentName - Name of the agent for error reporting
+ * @param action - Action being performed (e.g., "audio", "image", "movie")
+ * @param target - Target of the action (e.g., "audioFile", "imageFile")
+ * @param suppressError - If true, returns error object instead of throwing
+ * @param fn - Async function to execute
+ * @returns Result of fn or error object if suppressError is true
+ */
+export declare const wrapAgentError: <T>(agentName: string, action: string, target: string, suppressError: boolean | undefined, fn: () => Promise<T>) => Promise<T | {
+    error: unknown;
+}>;
+/**
+ * Validates that an API key is present, throwing an error if missing
+ *
+ * @param apiKey - API key to validate
+ * @param agentName - Name of the agent for error reporting
+ * @param action - Action being performed
+ * @param envVarName - Environment variable name for the API key
+ * @returns The validated API key
+ * @throws Error if API key is missing
+ */
+export declare const requireApiKey: (apiKey: string | undefined, agentName: string, action: string, envVarName: string) => string;

package/lib/utils/agent_error_handler.js

@@ -0,0 +1,59 @@
+import { GraphAILogger } from "graphai";
+import { AuthenticationError, RateLimitError } from "openai";
+import { agentIncorrectAPIKeyError, agentAPIRateLimitError, agentGenerationError, hasCause, } from "./error_cause.js";
+/**
+ * Wraps agent execution with standardized error handling
+ *
+ * @param agentName - Name of the agent for error reporting
+ * @param action - Action being performed (e.g., "audio", "image", "movie")
+ * @param target - Target of the action (e.g., "audioFile", "imageFile")
+ * @param suppressError - If true, returns error object instead of throwing
+ * @param fn - Async function to execute
+ * @returns Result of fn or error object if suppressError is true
+ */
+export const wrapAgentError = async (agentName, action, target, suppressError, fn) => {
+    try {
+        return await fn();
+    }
+    catch (error) {
+        if (suppressError) {
+            return { error };
+        }
+        GraphAILogger.info(error);
+        if (error instanceof AuthenticationError) {
+            throw new Error(`Failed: 401 Incorrect API key provided`, {
+                cause: agentIncorrectAPIKeyError(agentName, action, target),
+            });
+        }
+        if (error instanceof RateLimitError) {
+            throw new Error("You exceeded your current quota, please check your plan and billing details", {
+                cause: agentAPIRateLimitError(agentName, action, target),
+            });
+        }
+        if (hasCause(error)) {
+            throw error;
+        }
+        throw new Error(`Failed with ${agentName}`, {
+            cause: agentGenerationError(agentName, action, target),
+        });
+    }
+};
+/**
+ * Validates that an API key is present, throwing an error if missing
+ *
+ * @param apiKey - API key to validate
+ * @param agentName - Name of the agent for error reporting
+ * @param action - Action being performed
+ * @param envVarName - Environment variable name for the API key
+ * @returns The validated API key
+ * @throws Error if API key is missing
+ */
+export const requireApiKey = (apiKey, agentName, action, envVarName) => {
+    if (!apiKey) {
+        const { apiKeyMissingError } = require("./error_cause.js");
+        throw new Error(`${envVarName} is required`, {
+            cause: apiKeyMissingError(agentName, action, envVarName),
+        });
+    }
+    return apiKey;
+};

package/lib/utils/agent_info_factory.d.ts

@@ -0,0 +1,12 @@
+import type { AgentFunction, AgentFunctionInfo } from "graphai";
+/**
+ * Creates standardized AgentFunctionInfo metadata for agents
+ *
+ * @param name - Agent name (e.g., "ttsOpenaiAgent")
+ * @param agent - Agent function implementation
+ * @param category - Agent categories (e.g., ["tts"], ["image"])
+ * @param description - Human-readable description
+ * @param environmentVariables - Required environment variables (e.g., ["OPENAI_API_KEY"])
+ * @returns AgentFunctionInfo with standardized metadata
+ */
+export declare const createAgentInfo: <P = unknown, R = unknown, I = unknown, C = unknown>(name: string, agent: AgentFunction<P, R, I, C>, category: string[], description: string, environmentVariables?: string[]) => AgentFunctionInfo;

package/lib/utils/agent_info_factory.js

@@ -0,0 +1,23 @@
+import { PROJECT_METADATA } from "./const.js";
+/**
+ * Creates standardized AgentFunctionInfo metadata for agents
+ *
+ * @param name - Agent name (e.g., "ttsOpenaiAgent")
+ * @param agent - Agent function implementation
+ * @param category - Agent categories (e.g., ["tts"], ["image"])
+ * @param description - Human-readable description
+ * @param environmentVariables - Required environment variables (e.g., ["OPENAI_API_KEY"])
+ * @returns AgentFunctionInfo with standardized metadata
+ */
+export const createAgentInfo = (name, agent, category, description, environmentVariables = []) => ({
+    name,
+    agent,
+    mock: agent,
+    samples: [],
+    description,
+    category,
+    author: PROJECT_METADATA.author,
+    repository: PROJECT_METADATA.repository,
+    license: PROJECT_METADATA.license,
+    environmentVariables,
+});

package/lib/utils/context.d.ts

@@ -255,13 +255,13 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
         movieParams?: {
             provider?: string | undefined;
             model?: string | undefined;
+            fillOption?: {
+                style: "aspectFit" | "aspectFill";
+            } | undefined;
             transition?: {
                 type: "fade" | "slideout_left";
                 duration: number;
             } | undefined;
-            fillOption?: {
-                style: "aspectFit" | "aspectFill";
-            } | undefined;
         } | undefined;
         lipSyncParams?: {
             provider?: string | undefined;
@@ -570,13 +570,13 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
             movieParams?: {
                 provider?: string | undefined;
                 model?: string | undefined;
+                fillOption?: {
+                    style: "aspectFit" | "aspectFill";
+                } | undefined;
                 transition?: {
                     type: "fade" | "slideout_left";
                     duration: number;
                 } | undefined;
-                fillOption?: {
-                    style: "aspectFit" | "aspectFill";
-                } | undefined;
             } | undefined;
             lipSyncParams?: {
                 provider?: string | undefined;
@@ -719,13 +719,13 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
         movieParams?: {
             provider?: string | undefined;
             model?: string | undefined;
+            fillOption?: {
+                style: "aspectFit" | "aspectFill";
+            } | undefined;
             transition?: {
                 type: "fade" | "slideout_left";
                 duration: number;
             } | undefined;
-            fillOption?: {
-                style: "aspectFit" | "aspectFill";
-            } | undefined;
         } | undefined;
         lipSyncParams?: {
             provider?: string | undefined;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "2.0.0",
+  "version": "2.0.2",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -74,7 +74,7 @@
   "homepage": "https://github.com/receptron/mulmocast-cli#readme",
   "dependencies": {
     "@google-cloud/text-to-speech": "^6.4.0",
-    "@google/genai": "^1.29.0",
+    "@google/genai": "^1.29.1",
     "@graphai/anthropic_agent": "^2.0.11",
     "@graphai/browserless_agent": "^2.0.1",
     "@graphai/gemini_agent": "^2.0.1",
@@ -84,9 +84,9 @@
     "@graphai/stream_agent_filter": "^2.0.2",
     "@graphai/vanilla": "^2.0.12",
     "@graphai/vanilla_node_agents": "^2.0.4",
-    "@inquirer/input": "^4.2.5",
-    "@inquirer/select": "^4.4.0",
-    "@modelcontextprotocol/sdk": "^1.21.1",
+    "@inquirer/input": "^4.3.0",
+    "@inquirer/select": "^4.4.1",
+    "@modelcontextprotocol/sdk": "^1.22.0",
     "@mozilla/readability": "^0.6.0",
     "@tavily/core": "^0.5.11",
     "archiver": "^7.0.1",
@@ -94,11 +94,11 @@
     "dotenv": "^17.2.3",
     "fluent-ffmpeg": "^2.1.3",
     "graphai": "^2.0.16",
-    "jsdom": "^27.1.0",
+    "jsdom": "^27.2.0",
     "marked": "^17.0.0",
     "mulmocast-vision": "^1.0.4",
     "ora": "^9.0.0",
-    "puppeteer": "^24.29.1",
+    "puppeteer": "^24.30.0",
     "replicate": "^1.3.1",
     "yaml": "^2.8.1",
     "yargs": "^18.0.0",
@@ -109,7 +109,7 @@
     "@types/archiver": "^7.0.0",
     "@types/fluent-ffmpeg": "^2.1.28",
     "@types/jsdom": "^27.0.0",
-    "@types/yargs": "^17.0.34",
+    "@types/yargs": "^17.0.35",
     "eslint": "^9.39.1",
     "eslint-config-prettier": "^10.1.8",
     "eslint-plugin-prettier": "^5.5.4",
@@ -117,7 +117,7 @@
     "prettier": "^3.6.2",
     "tsx": "^4.20.6",
     "typescript": "^5.9.3",
-    "typescript-eslint": "^8.46.3"
+    "typescript-eslint": "^8.46.4"
   },
   "engines": {
     "node": ">=20.0.0"

package/scripts/test/test_movie2.json

@@ -0,0 +1,56 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "lang": "en",
+  "movieParams": {
+    "provider": "google"
+  },
+  "lipSyncParams": {
+    "provider": "replicate",
+    "model": "bytedance/latentsync"
+  },
+  "canvasSize": {
+    "width": 1280,
+    "height": 720
+  },
+  "beats": [
+    {
+      "text": "Hello, I'm macoro",
+      "image": {
+        "type": "image",
+        "source": {
+          "kind": "url",
+          "url": "https://raw.githubusercontent.com/receptron/mulmocast-media/refs/heads/main/characters/macoro.png"
+        }
+      },
+      "moviePrompt": "Jump and bounce",
+      "duration": 5
+    },
+    {
+      "text": "こんにちは、マコロです",
+      "image": {
+        "type": "image",
+        "source": {
+          "kind": "url",
+          "url": "https://raw.githubusercontent.com/receptron/mulmocast-media/refs/heads/main/characters/macoro.png"
+        }
+      },
+      "enableLipSync": true,
+      "duration": 5
+    },
+    {
+      "text": "こんにちは、マコロです",
+      "image": {
+        "type": "image",
+        "source": {
+          "kind": "url",
+          "url": "https://raw.githubusercontent.com/receptron/mulmocast-media/refs/heads/main/characters/explaining_concepts.png"
+        }
+      },
+      "moviePrompt": "Jump and bounce",
+      "enableLipSync": true,
+      "duration": 5
+    }
+  ]
+}

package/scripts/test/test_movie2.json~

@@ -0,0 +1,40 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "lang": "en",
+  "movieParams": {
+    "provider": "google"
+  },
+  "canvasSize": {
+    "width": 720,
+    "height": 1280
+  },
+  "imageParams": {
+    "provider": "openai",
+    "model": "dall-e-3",
+    "style": "Photo realistic, cinematic style.",
+    "images": {
+      "optimus": {
+        "type": "image",
+        "source": {
+          "kind": "url",
+          "url": "https://raw.githubusercontent.com/receptron/mulmocast-media/refs/heads/main/characters/optimus.png"
+        }
+      }
+    }
+  },
+  "beats": [
+    {
+      "text": "Image with both image and movie prompt",
+      "imagePrompt": "巨大なロケット「Starship」が発射台から打ち上がろうとしている様子。空には星と火星が浮かんでいる。",
+      "moviePrompt": "Startship lifts off.",
+      "duration": 5
+    },
+    {
+      "text": "Image with only movie prompt",
+      "moviePrompt": "Startship lifts off.",
+      "duration": 5
+    }
+  ]
+}

package/scripts/test/test_play_to_end.json~

@@ -0,0 +1,65 @@
+{
+  "$mulmocast": {
+    "version": "1.1",
+    "credit": "closing"
+  },
+  "canvasSize": {
+    "width": 1536,
+    "height": 1024
+  },
+  "speechParams": {
+    "speakers": {
+      "Presenter": {
+        "displayName": {
+          "en": "Presenter"
+        },
+        "voiceId": "shimmer"
+      }
+    }
+  },
+  "imageParams": {
+    "provider": "openai",
+    "style": "<style>Vibrant 3D animation style inspired by K-pop aesthetics, with glossy, stylized characters. The overall visual style combines elements of modern animation, game cinematics, and fashion-forward character design, with sleek outlines, glowing effects, and a polished, cinematic finish.</style>",
+    "images": {
+      "min": {
+        "type": "image",
+        "source": {
+          "kind": "url",
+          "url": "https://raw.githubusercontent.com/receptron/mulmocast-media/refs/heads/main/characters/min_anime.png"
+        }
+      }
+    }
+  },
+  "movieParams": {
+    "provider": "replicate"
+  },
+  "soundEffectParams": {
+    "provider": "replicate"
+  },
+  "captionParams": {
+    "lang": "en",
+    "styles": ["font-size: 64px", "width: 90%", "padding-left: 5%", "padding-right: 5%"]
+  },
+  "audioParams": {
+    "padding": 0,
+    "introPadding": 0,
+    "closingPadding": 0,
+    "outroPadding": 0,
+    "bgm": {
+      "kind": "url",
+      "url": "https://raw.githubusercontent.com/receptron/mulmocast-media/refs/heads/main/music/finetuning_with_you.mp3"
+    },
+    "bgmVolume": 1,
+    "audioVolume": 0,
+    "suppressSpeech": true
+  },
+  "title": "Music Video",
+  "lang": "en",
+  "beats": [
+    {
+      "text": "Finetuning with you",
+      "moviePrompt": "The singer preparing to sing a song.",
+      "playGenMovieToEnd": true
+    }
+  ]
+}