mulmocast 0.0.9 → 0.0.11

package/lib/types/schema.js

@@ -134,7 +134,6 @@ const imageIdSchema = z.string();
 export const mulmoImageParamsSchema = z
     .object({
     model: z.string().optional(), // default: provider specific
-    size: z.string().optional(), // default: provider specific
     style: z.string().optional(), // optional image style
     moderation: z.string().optional(), // optional image style
     images: z.record(imageIdSchema, mulmoImageMediaSchema).optional(),
@@ -170,6 +169,7 @@ export const mulmoBeatSchema = z
     .object({
     speaker: speakerIdSchema.default("Presenter"),
     text: z.string().describe("Text to be spoken. If empty, the audio is not generated."),
+    description: z.string().optional(),
     image: mulmoImageAssetSchema.optional(),
     audio: mulmoAudioAssetSchema.optional(),
     duration: z.number().optional().describe("Duration of the beat. Used only when the text is empty"),
@@ -178,7 +178,8 @@ export const mulmoBeatSchema = z
     speechOptions: speechOptionsSchema.optional(),
     textSlideParams: textSlideParamsSchema.optional(),
     imageNames: z.array(imageIdSchema).optional(), // list of image names to use for image generation. The default is all images in the imageParams.images.
-    imagePrompt: z.string().optional(), // specified or inserted by preprocessor
+    imagePrompt: z.string().optional(),
+    moviePrompt: z.string().optional(),
 })
     .strict();
 export const mulmoCanvasDimensionSchema = z
@@ -202,6 +203,13 @@ export const mulmoSpeechParamsSchema = z
 })
     .strict();
 export const text2ImageProviderSchema = z.union([z.literal("openai"), z.literal("google")]).default("openai");
+export const text2MovieProviderSchema = z.union([z.literal("openai"), z.literal("google")]).default("google");
+export const mulmoMovieParamsSchema = z
+    .object({
+    provider: text2MovieProviderSchema.optional(),
+    model: z.string().optional(), // default: provider specific
+})
+    .strict();
 export const mulmoPresentationStyleSchema = z.object({
     $mulmocast: mulmoCastCreditSchema,
     canvasSize: mulmoCanvasDimensionSchema, // has default value
@@ -220,6 +228,7 @@ export const mulmoPresentationStyleSchema = z.object({
         provider: text2ImageProviderSchema, // has default value
     })
         .optional(),
+    movieParams: mulmoMovieParamsSchema.optional(),
     // for textSlides
     textSlideParams: textSlideParamsSchema.optional(),
     // videoParams: videoParamsSchema.optional(),
@@ -257,6 +266,7 @@ export const mulmoStudioBeatSchema = z
     duration: z.number().optional(),
     audioFile: z.string().optional(),
     imageFile: z.string().optional(), // path to the image
+    movieFile: z.string().optional(), // path to the movie file
     captionFile: z.string().optional(), // path to the caption image
 })
     .strict();
@@ -276,6 +286,7 @@ export const mulmoSessionStateSchema = z.object({
     inBeatSession: z.object({
         audio: z.set(z.number()),
         image: z.set(z.number()),
+        movie: z.set(z.number()),
         multiLingual: z.set(z.number()),
         caption: z.set(z.number()),
     }),
@@ -298,6 +309,7 @@ export const mulmoStudioSchema = z
         inBeatSession: {
             audio: new Set(),
             image: new Set(),
+            movie: new Set(),
             multiLingual: new Set(),
             caption: new Set(),
         },

package/lib/types/type.d.ts

@@ -1,4 +1,4 @@
-import { langSchema, localizedTextSchema, mulmoBeatSchema, mulmoScriptSchema, mulmoStudioSchema, mulmoStudioBeatSchema, mulmoStoryboardSchema, mulmoStoryboardSceneSchema, mulmoStudioMultiLingualSchema, mulmoStudioMultiLingualDataSchema, speakerDictionarySchema, mulmoImageParamsSchema, mulmoSpeechParamsSchema, textSlideParamsSchema, speechOptionsSchema, mulmoCanvasDimensionSchema, mulmoScriptTemplateSchema, text2ImageProviderSchema, text2SpeechProviderSchema, mulmoPresentationStyleSchema, multiLingualTextsSchema, mulmoMermaidMediaSchema, mulmoTextSlideMediaSchema, mulmoMarkdownMediaSchema, mulmoImageMediaSchema, mulmoChartMediaSchema, mediaSourceSchema } from "./schema.js";
+import { langSchema, localizedTextSchema, mulmoBeatSchema, mulmoScriptSchema, mulmoStudioSchema, mulmoStudioBeatSchema, mulmoStoryboardSchema, mulmoStoryboardSceneSchema, mulmoStudioMultiLingualSchema, mulmoStudioMultiLingualDataSchema, speakerDictionarySchema, mulmoImageParamsSchema, mulmoMovieParamsSchema, mulmoSpeechParamsSchema, textSlideParamsSchema, speechOptionsSchema, mulmoCanvasDimensionSchema, mulmoScriptTemplateSchema, text2ImageProviderSchema, text2MovieProviderSchema, text2SpeechProviderSchema, mulmoPresentationStyleSchema, multiLingualTextsSchema, mulmoMermaidMediaSchema, mulmoTextSlideMediaSchema, mulmoMarkdownMediaSchema, mulmoImageMediaSchema, mulmoChartMediaSchema, mediaSourceSchema } from "./schema.js";
 import { pdf_modes, pdf_sizes, storyToScriptGenerateMode } from "../utils/const.js";
 import { LLM } from "../utils/utils.js";
 import { z } from "zod";
@@ -10,6 +10,7 @@ export type SpeechOptions = z.infer<typeof speechOptionsSchema>;
 export type MulmoImageParams = z.infer<typeof mulmoImageParamsSchema>;
 export type TextSlideParams = z.infer<typeof textSlideParamsSchema>;
 export type Text2ImageProvider = z.infer<typeof text2ImageProviderSchema>;
+export type Text2MovieProvider = z.infer<typeof text2MovieProviderSchema>;
 export type Text2SpeechProvider = z.infer<typeof text2SpeechProviderSchema>;
 export type LocalizedText = z.infer<typeof localizedTextSchema>;
 export type MulmoScript = z.infer<typeof mulmoScriptSchema>;
@@ -24,6 +25,7 @@ export type MulmoScriptTemplate = z.infer<typeof mulmoScriptTemplateSchema>;
 export type MulmoStudioMultiLingual = z.infer<typeof mulmoStudioMultiLingualSchema>;
 export type MulmoStudioMultiLingualData = z.infer<typeof mulmoStudioMultiLingualDataSchema>;
 export type MultiLingualTexts = z.infer<typeof multiLingualTextsSchema>;
+export type MulmoMovieParams = z.infer<typeof mulmoMovieParamsSchema>;
 export type MulmoTextSlideMedia = z.infer<typeof mulmoTextSlideMediaSchema>;
 export type MulmoMarkdownMedia = z.infer<typeof mulmoMarkdownMediaSchema>;
 export type MulmoImageMedia = z.infer<typeof mulmoImageMediaSchema>;

package/lib/utils/file.d.ts

@@ -1,6 +1,7 @@
 import { MulmoScript, MulmoScriptTemplate, MulmoMediaSource, MulmoStudioContext } from "../types/index.js";
 import { PDFMode } from "../types/index.js";
 import { ZodSchema } from "zod";
+export declare const updateNpmRoot: (_npmRoot: string) => void;
 export declare function readMulmoScriptFile<T = MulmoScript>(path: string, errorMessage: string): {
     mulmoData: T;
     mulmoDataPath: string;

package/lib/utils/file.js

@@ -8,6 +8,10 @@ import { MulmoStudioContextMethods } from "../methods/index.js";
 import { mulmoScriptTemplateSchema } from "../types/schema.js";
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = path.dirname(__filename);
+let npmRoot = path.resolve(__dirname, "../../");
+export const updateNpmRoot = (_npmRoot) => {
+    npmRoot = _npmRoot;
+};
 export function readMulmoScriptFile(arg2, errorMessage) {
     const scriptPath = path.resolve(arg2);
     if (!fs.existsSync(scriptPath)) {
@@ -85,7 +89,7 @@ export const getOutputPdfFilePath = (outDirPath, fileName, pdfMode, lang) => {
     return path.resolve(outDirPath, `${fileName}_${pdfMode}.pdf`);
 };
 export const getTemplateFilePath = (templateName) => {
-    return path.resolve(__dirname, "../../assets/templates/" + templateName + ".json");
+    return path.resolve(npmRoot, "./assets/templates/" + templateName + ".json");
 };
 export const mkdir = (dirPath) => {
     if (!fs.existsSync(dirPath)) {
@@ -93,12 +97,12 @@ export const mkdir = (dirPath) => {
         fs.mkdirSync(dirPath, { recursive: true });
     }
 };
-export const silentPath = path.resolve(__dirname, "../../assets/audio/silent300.mp3");
-export const silentLastPath = path.resolve(__dirname, "../../assets/audio/silent800.mp3");
-export const silent60secPath = path.resolve(__dirname, "../../assets/audio/silent60sec.mp3");
-export const defaultBGMPath = path.resolve(__dirname, "../../assets/music/StarsBeyondEx.mp3");
+export const silentPath = path.resolve(npmRoot, "./assets/audio/silent300.mp3");
+export const silentLastPath = path.resolve(npmRoot, "./assets/audio/silent800.mp3");
+export const silent60secPath = path.resolve(npmRoot, "./assets/audio/silent60sec.mp3");
+export const defaultBGMPath = path.resolve(npmRoot, "./assets/music/StarsBeyondEx.mp3");
 export const getHTMLFile = (filename) => {
-    const htmlPath = path.resolve(__dirname, `../../assets/html/${filename}.html`);
+    const htmlPath = path.resolve(npmRoot, `./assets/html/${filename}.html`);
     return fs.readFileSync(htmlPath, "utf-8");
 };
 // for cli
@@ -121,7 +125,7 @@ export const getFullPath = (baseDirPath, file) => {
     return path.resolve(file);
 };
 export const readScriptTemplateFile = (scriptName) => {
-    const scriptPath = path.resolve(__dirname, "../../scripts/templates", scriptName);
+    const scriptPath = path.resolve(npmRoot, "./scripts/templates", scriptName);
     const scriptData = fs.readFileSync(scriptPath, "utf-8");
     // NOTE: We don't want to schema parse the script here to eliminate default values.
     return JSON.parse(scriptData);
@@ -142,7 +146,7 @@ export const readTemplatePrompt = (templateName) => {
     return prompt;
 };
 export const getAvailableTemplates = () => {
-    const templatesDir = path.resolve(__dirname, "../../assets/templates");
+    const templatesDir = path.resolve(npmRoot, "./assets/templates");
     if (!fs.existsSync(templatesDir)) {
         return [];
     }

package/lib/utils/image_plugins/image.d.ts

@@ -1,2 +1,2 @@
 export declare const imageType = "image";
-export declare const process: (params: import("../../types/type.js").ImageProcessorParams) => string | undefined;
+export declare const process: (params: import("../../index.js").ImageProcessorParams) => string | undefined;

package/lib/utils/image_plugins/movie.d.ts

@@ -1,2 +1,2 @@
 export declare const imageType = "movie";
-export declare const process: (params: import("../../types/type.js").ImageProcessorParams) => string | undefined;
+export declare const process: (params: import("../../index.js").ImageProcessorParams) => string | undefined;

package/lib/utils/preprocess.d.ts

@@ -5,6 +5,7 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
         hash?: string | undefined;
         audioFile?: string | undefined;
         imageFile?: string | undefined;
+        movieFile?: string | undefined;
         captionFile?: string | undefined;
     }[];
     multiLingual: {
@@ -168,9 +169,9 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
                 type: "midi";
                 source: string;
             } | undefined;
+            description?: string | undefined;
             imageParams?: {
                 model?: string | undefined;
-                size?: string | undefined;
                 style?: string | undefined;
                 moderation?: string | undefined;
                 images?: Record<string, {
@@ -198,13 +199,14 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
             } | undefined;
             imageNames?: string[] | undefined;
             imagePrompt?: string | undefined;
+            moviePrompt?: string | undefined;
         }[];
         lang?: string | undefined;
         title?: string | undefined;
+        description?: string | undefined;
         imageParams?: {
             provider: "openai" | "google";
             model?: string | undefined;
-            size?: string | undefined;
             style?: string | undefined;
             moderation?: string | undefined;
             images?: Record<string, {
@@ -227,8 +229,11 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
         textSlideParams?: {
             cssStyles: string | string[];
         } | undefined;
+        movieParams?: {
+            model?: string | undefined;
+            provider?: "openai" | "google" | undefined;
+        } | undefined;
         omitCaptions?: boolean | undefined;
-        description?: string | undefined;
         references?: {
             type: "image" | "audio" | "article" | "paper" | "video";
             url: string;
@@ -250,6 +255,7 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
         };
         inBeatSession: {
             image: Set<number>;
+            movie: Set<number>;
             audio: Set<number>;
             multiLingual: Set<number>;
             caption: Set<number>;

package/lib/utils/utils.d.ts

@@ -15,3 +15,4 @@ export declare const chunkArray: <T>(array: T[], size?: number) => T[][];
 export declare const isHttp: (fileOrUrl: string) => boolean;
 export declare const text2hash: (input: string) => string;
 export declare const localizedText: (beat: MulmoBeat, multiLingualData?: MulmoStudioMultiLingualData, lang?: string) => string;
+export declare const sleep: (milliseconds: number) => Promise<unknown>;

package/lib/utils/utils.js

@@ -52,3 +52,6 @@ export const localizedText = (beat, multiLingualData, lang) => {
     }
     return beat.text;
 };
+export const sleep = async (milliseconds) => {
+    return await new Promise((resolve) => setTimeout(resolve, milliseconds));
+};

package/package.json

@@ -1,9 +1,9 @@
 {
   "name": "mulmocast",
-  "version": "0.0.9",
+  "version": "0.0.11",
   "description": "",
   "type": "module",
-  "main": "lib/cli/bin.js",
+  "main": "lib/index.js",
   "bin": {
     "mulmo": "lib/cli/bin.js"
   },
@@ -69,27 +69,27 @@
     "google-auth-library": "^9.15.1",
     "graphai": "^2.0.5",
     "inquirer": "^12.6.1",
-    "marked": "^15.0.11",
+    "marked": "^15.0.12",
     "ora": "^8.2.0",
     "pdf-lib": "^1.17.1",
-    "puppeteer": "^24.8.1",
+    "puppeteer": "^24.10.0",
     "yaml": "^2.8.0",
     "yargs": "^17.7.2",
-    "zod": "^3.24.4",
+    "zod": "^3.25.51",
     "zod-to-json-schema": "^3.24.5"
   },
   "devDependencies": {
     "@receptron/test_utils": "^2.0.0",
     "@types/fluent-ffmpeg": "^2.1.26",
     "@types/yargs": "^17.0.33",
-    "eslint": "^9.27.0",
+    "eslint": "^9.28.0",
     "eslint-config-prettier": "^10.1.5",
-    "eslint-plugin-prettier": "^5.4.0",
+    "eslint-plugin-prettier": "^5.4.1",
     "prettier": "^3.3.3",
     "ts-node": "^10.9.2",
     "tsx": "^4.19.4",
     "typescript": "^5.7.3",
-    "typescript-eslint": "^8.32.1"
+    "typescript-eslint": "^8.33.1"
   },
   "engines": {
     "node": ">=16.0.0"

package/scripts/templates/movie_prompts_template.json

@@ -0,0 +1,50 @@
+{
+  "$mulmocast": {
+    "version": "1.0",
+    "credit": "closing"
+  },
+  "title": "[TITLE: Brief, engaging title for the topic]",
+  "lang": "en",
+  "references": [
+    {
+      "url": "[SOURCE_URL: URL of the source material]",
+      "title": "[SOURCE_TITLE: Title of the referenced article, or paper]",
+      "type": "[SOURCE_TYPE: article, paper]"
+    }
+  ],
+  "movieParams": {
+    "provider": "google"
+  },
+  "beats": [
+    {
+      "text": "[OPENING_BEAT: Introduce the topic with a hook. Reference the source material and set up why this topic matters. Usually 2-3 sentences that grab attention and provide context.]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[MAIN_CONCEPT: Define or explain the core concept/idea. This should be the central focus of your narrative. Keep it clear and accessible.]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[SUPPORTING_DETAIL_1: Additional context, examples, or elaboration that helps illustrate the main concept. This could include how it works, why it's important, or real-world applications.]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[SUPPORTING_DETAIL_2: Continue with more examples, deeper explanation, or different aspects of the topic if needed.]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[ADDITIONAL_BEATS: Add more beats as necessary to fully explore the topic. Complex topics may require 6-10+ beats to cover adequately. Each beat should advance the narrative or provide valuable information.]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[CONCLUSION/IMPACT: Wrap up with the significance, implications, or key takeaway. Help the audience understand why this matters to them.]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    }
+  ]
+}

package/scripts/templates/shorts_template.json

@@ -0,0 +1,52 @@
+{
+  "$mulmocast": {
+    "version": "1.0"
+  },
+  "title": "[TITLE: Brief, engaging title for the topic]",
+  "lang": "en",
+  "movieParams": {
+    "provider": "google"
+  },
+  "beats": [
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    },
+    {
+      "text": "[NARRATION: Short narration for the beat. Up to 20 words]",
+      "imagePrompt": "[IMAGE_PROMPT: A prompt for the image to be generated for this beat.]",
+      "moviePrompt": "[MOVIE_PROMPT: A movie prompt for that image.]"
+    }
+  ]
+}