mulmocast 0.0.26 → 0.0.27

package/README.md

@@ -173,7 +173,7 @@ DEFAULT_OPENAI_IMAGE_MODEL=gpt-image-1 # required for high-quality Ghibli-style
 
 ### Step 2: Generate a Ghibli-style MulmoScript
 ```bash
-mulmo tool scripting -i -t ghibli_strips -o ./ -s story
+mulmo tool scripting -i -t ghibli_comic -o ./ -s story
 ```
 This will initiate an interactive conversation with the AI to create your Ghibli-inspired story. Once completed, a JSON file (e.g., `story-1747834931950.json`) will be generated.
 

package/lib/actions/audio.js

@@ -32,7 +32,7 @@ const getAudioPath = (context, beat, audioFile) => {
         }
         throw new Error("Invalid audio source");
     }
-    if (beat.text === undefined || beat.text === "") {
+    if (beat.text === undefined || beat.text === "" || context.studio.script.audioParams.suppressSpeech) {
         return undefined; // It indicates that the audio is not needed.
     }
     return audioFile;

package/lib/actions/images.js

@@ -372,13 +372,11 @@ const prepareGenerateImages = async (context) => {
     const imageProjectDirPath = MulmoStudioContextMethods.getImageProjectDirPath(context);
     const outDirPath = MulmoStudioContextMethods.getOutDirPath(context);
     mkdir(imageProjectDirPath);
-    const imageAgentInfo = MulmoPresentationStyleMethods.getImageAgentInfo(context.presentationStyle, context.dryRun);
+    const imageAgentInfo = MulmoPresentationStyleMethods.getImageAgentInfo(context.presentationStyle);
     const htmlImageAgentInfo = MulmoPresentationStyleMethods.getHtmlImageAgentInfo(context.presentationStyle);
     const imageRefs = await getImageRefs(context);
     // Determine movie agent based on provider
     const getMovieAgent = () => {
-        if (context.dryRun)
-            return "mediaMockAgent";
         const provider = context.presentationStyle.movieParams?.provider ?? "google";
         switch (provider) {
             case "replicate":

package/lib/cli/commands/audio/builder.d.ts

@@ -7,8 +7,6 @@ export declare const builder: (yargs: Argv) => Argv<{
     l: string | undefined;
 } & {
     f: boolean;
-} & {
-    dryRun: boolean;
 } & {
     p: string | undefined;
 } & {

package/lib/cli/commands/image/builder.d.ts

@@ -7,8 +7,6 @@ export declare const builder: (yargs: Argv) => Argv<{
     l: string | undefined;
 } & {
     f: boolean;
-} & {
-    dryRun: boolean;
 } & {
     p: string | undefined;
 } & {

package/lib/cli/commands/movie/builder.d.ts

@@ -7,8 +7,6 @@ export declare const builder: (yargs: Argv) => Argv<{
     l: string | undefined;
 } & {
     f: boolean;
-} & {
-    dryRun: boolean;
 } & {
     p: string | undefined;
 } & {

package/lib/cli/commands/pdf/builder.d.ts

@@ -7,8 +7,6 @@ export declare const builder: (yargs: Argv) => Argv<{
     l: string | undefined;
 } & {
     f: boolean;
-} & {
-    dryRun: boolean;
 } & {
     p: string | undefined;
 } & {

package/lib/cli/commands/translate/builder.d.ts

@@ -7,8 +7,6 @@ export declare const builder: (yargs: Argv) => Argv<import("yargs").Omit<{
     l: string | undefined;
 } & {
     f: boolean;
-} & {
-    dryRun: boolean;
 } & {
     p: string | undefined;
 } & {

package/lib/cli/common.d.ts

@@ -7,8 +7,6 @@ export declare const commonOptions: (yargs: Argv) => Argv<{
     l: string | undefined;
 } & {
     f: boolean;
-} & {
-    dryRun: boolean;
 } & {
     p: string | undefined;
 } & {

package/lib/cli/common.js

@@ -25,11 +25,6 @@ export const commonOptions = (yargs) => {
         describe: "Force regenerate",
         type: "boolean",
         default: false,
-    })
-        .option("dryRun", {
-        describe: "Dry run",
-        type: "boolean",
-        default: false,
     })
         .option("p", {
         alias: "presentationStyle",

package/lib/cli/helpers.d.ts

@@ -1,20 +1,10 @@
-import type { MulmoScript, MulmoStudioContext, MulmoPresentationStyle, MulmoStudioMultiLingual } from "../types/type.js";
 import type { CliArgs } from "../types/cli_types.js";
+import { FileObject, InitOptions, MulmoStudioContext } from "../types/index.js";
+export declare const runTranslateIfNeeded: (context: MulmoStudioContext, argv: {
+    l?: string;
+    c?: string;
+}) => Promise<void>;
 export declare const setGraphAILogger: (verbose: boolean | undefined, logValues?: Record<string, unknown>) => void;
-export interface FileObject {
-    baseDirPath: string;
-    mulmoFilePath: string;
-    mulmoFileDirPath: string;
-    outDirPath: string;
-    imageDirPath: string;
-    audioDirPath: string;
-    isHttpPath: boolean;
-    fileOrUrl: string;
-    outputStudioFilePath: string;
-    outputMultilingualFilePath: string;
-    presentationStylePath: string | undefined;
-    fileName: string;
-}
 export declare const getFileObject: (args: {
     basedir?: string;
     outdir?: string;
@@ -23,22 +13,4 @@ export declare const getFileObject: (args: {
     presentationStyle?: string;
     file: string;
 }) => FileObject;
-export declare const fetchScript: (isHttpPath: boolean, mulmoFilePath: string, fileOrUrl: string) => Promise<MulmoScript | null>;
-export declare const getMultiLingual: (multilingualFilePath: string, beatsLength: number) => MulmoStudioMultiLingual;
-export declare const getPresentationStyle: (presentationStylePath: string | undefined) => MulmoPresentationStyle | null;
-type InitOptions = {
-    b?: string;
-    o?: string;
-    i?: string;
-    a?: string;
-    file?: string;
-    l?: string;
-    c?: string;
-    p?: string;
-};
-export declare const initializeContext: (argv: CliArgs<InitOptions>) => Promise<MulmoStudioContext | null>;
-export declare const runTranslateIfNeeded: (context: MulmoStudioContext, argv: {
-    l?: string;
-    c?: string;
-}) => Promise<void>;
-export {};
+export declare const initializeContext: (argv: CliArgs<InitOptions>, raiseError?: boolean) => Promise<MulmoStudioContext | null>;

package/lib/cli/helpers.js

@@ -2,12 +2,17 @@ import { GraphAILogger } from "graphai";
 import fs from "fs";
 import path from "path";
 import clipboardy from "clipboardy";
-import { getBaseDirPath, getFullPath, readMulmoScriptFile, fetchMulmoScriptFile, getOutputStudioFilePath, resolveDirPath, mkdir, getOutputMultilingualFilePath, generateTimestampedFileName, } from "../utils/file.js";
+import { getBaseDirPath, getFullPath, getOutputStudioFilePath, resolveDirPath, mkdir, getOutputMultilingualFilePath, generateTimestampedFileName, } from "../utils/file.js";
 import { isHttp } from "../utils/utils.js";
-import { createOrUpdateStudioData } from "../utils/preprocess.js";
 import { outDirName, imageDirName, audioDirName } from "../utils/const.js";
 import { translate } from "../actions/translate.js";
-import { mulmoCaptionParamsSchema, mulmoPresentationStyleSchema, mulmoStudioMultiLingualSchema } from "../types/schema.js";
+import { initializeContextFromFiles } from "../utils/context.js";
+export const runTranslateIfNeeded = async (context, argv) => {
+    if (argv.l || context.studio.script.captionParams?.lang) {
+        GraphAILogger.log("run translate");
+        await translate(context);
+    }
+};
 export const setGraphAILogger = (verbose, logValues) => {
     if (verbose) {
         if (logValues) {
@@ -63,44 +68,7 @@ export const getFileObject = (args) => {
         fileName,
     };
 };
-export const fetchScript = async (isHttpPath, mulmoFilePath, fileOrUrl) => {
-    if (isHttpPath) {
-        const res = await fetchMulmoScriptFile(fileOrUrl);
-        if (!res.result || !res.script) {
-            GraphAILogger.info(`ERROR: HTTP error! ${res.status} ${fileOrUrl}`);
-            return null;
-        }
-        return res.script;
-    }
-    if (!fs.existsSync(mulmoFilePath)) {
-        GraphAILogger.info(`ERROR: File not exists ${mulmoFilePath}`);
-        return null;
-    }
-    return readMulmoScriptFile(mulmoFilePath, "ERROR: File does not exist " + mulmoFilePath)?.mulmoData ?? null;
-};
-export const getMultiLingual = (multilingualFilePath, beatsLength) => {
-    if (fs.existsSync(multilingualFilePath)) {
-        const jsonData = readMulmoScriptFile(multilingualFilePath, "ERROR: File does not exist " + multilingualFilePath)?.mulmoData ?? null;
-        const dataSet = mulmoStudioMultiLingualSchema.parse(jsonData);
-        while (dataSet.length < beatsLength) {
-            dataSet.push({ multiLingualTexts: {} });
-        }
-        dataSet.length = beatsLength;
-        return dataSet;
-    }
-    return [...Array(beatsLength)].map(() => ({ multiLingualTexts: {} }));
-};
-export const getPresentationStyle = (presentationStylePath) => {
-    if (presentationStylePath) {
-        if (!fs.existsSync(presentationStylePath)) {
-            throw new Error(`ERROR: File not exists ${presentationStylePath}`);
-        }
-        const jsonData = readMulmoScriptFile(presentationStylePath, "ERROR: File does not exist " + presentationStylePath)?.mulmoData ?? null;
-        return mulmoPresentationStyleSchema.parse(jsonData);
-    }
-    return null;
-};
-export const initializeContext = async (argv) => {
+export const initializeContext = async (argv, raiseError = false) => {
     const files = getFileObject({
         basedir: argv.b,
         outdir: argv.o,
@@ -109,62 +77,6 @@ export const initializeContext = async (argv) => {
         presentationStyle: argv.p,
         file: argv.file ?? "",
     });
-    const { fileName, isHttpPath, fileOrUrl, mulmoFilePath, outputStudioFilePath, presentationStylePath, outputMultilingualFilePath } = files;
-    setGraphAILogger(argv.v, {
-        files,
-    });
-    const mulmoScript = await fetchScript(isHttpPath, mulmoFilePath, fileOrUrl);
-    if (!mulmoScript) {
-        return null;
-    }
-    const presentationStyle = getPresentationStyle(presentationStylePath);
-    // Create or update MulmoStudio file with MulmoScript
-    const currentStudio = readMulmoScriptFile(outputStudioFilePath);
-    try {
-        // validate mulmoStudioSchema. skip if __test_invalid__ is true
-        const studio = createOrUpdateStudioData(mulmoScript, currentStudio?.mulmoData, fileName);
-        const multiLingual = getMultiLingual(outputMultilingualFilePath, studio.beats.length);
-        if (argv.c) {
-            studio.script.captionParams = mulmoCaptionParamsSchema.parse({
-                ...(studio.script.captionParams ?? {}),
-                lang: argv.c,
-            });
-        }
-        return {
-            studio,
-            fileDirs: files,
-            force: Boolean(argv.f),
-            dryRun: Boolean(argv.dryRun),
-            lang: argv.l,
-            sessionState: {
-                inSession: {
-                    audio: false,
-                    image: false,
-                    video: false,
-                    multiLingual: false,
-                    caption: false,
-                    pdf: false,
-                },
-                inBeatSession: {
-                    audio: {},
-                    image: {},
-                    movie: {},
-                    multiLingual: {},
-                    caption: {},
-                },
-            },
-            presentationStyle: presentationStyle ?? studio.script,
-            multiLingual,
-        };
-    }
-    catch (error) {
-        GraphAILogger.info(`Error: invalid MulmoScript Schema: ${isHttpPath ? fileOrUrl : mulmoFilePath} \n ${error}`);
-        return null;
-    }
-};
-export const runTranslateIfNeeded = async (context, argv) => {
-    if (argv.l || context.studio.script.captionParams?.lang) {
-        GraphAILogger.log("run translate");
-        await translate(context);
-    }
+    setGraphAILogger(Boolean(argv.v), { files });
+    return await initializeContextFromFiles(files, raiseError, Boolean(argv.f), argv.c, argv.l);
 };

package/lib/methods/mulmo_presentation_style.d.ts

@@ -9,7 +9,7 @@ export declare const MulmoPresentationStyleMethods: {
     getSpeaker(presentationStyle: MulmoPresentationStyle, beat: MulmoBeat): SpeakerData;
     getProvider(presentationStyle: MulmoPresentationStyle, beat: MulmoBeat): Text2SpeechProvider;
     getVoiceId(presentationStyle: MulmoPresentationStyle, beat: MulmoBeat): string;
-    getImageAgentInfo(presentationStyle: MulmoPresentationStyle, dryRun?: boolean): Text2ImageAgentInfo;
+    getImageAgentInfo(presentationStyle: MulmoPresentationStyle): Text2ImageAgentInfo;
     getHtmlImageAgentInfo(presentationStyle: MulmoPresentationStyle): Text2HtmlAgentInfo;
     getImageType(_: MulmoPresentationStyle, beat: MulmoBeat): BeatMediaType;
 };

package/lib/methods/mulmo_presentation_style.js

@@ -57,7 +57,7 @@ export const MulmoPresentationStyleMethods = {
         const speaker = MulmoPresentationStyleMethods.getSpeaker(presentationStyle, beat);
         return speaker.voiceId;
     },
-    getImageAgentInfo(presentationStyle, dryRun = false) {
+    getImageAgentInfo(presentationStyle) {
         // Notice that we copy imageParams from presentationStyle and update
         // provider and model appropriately.
         const provider = text2ImageProviderSchema.parse(presentationStyle.imageParams?.provider);
@@ -66,7 +66,7 @@ export const MulmoPresentationStyleMethods = {
         };
         return {
             provider,
-            agent: dryRun ? "mediaMockAgent" : provider === "google" ? "imageGoogleAgent" : "imageOpenaiAgent",
+            agent: provider === "google" ? "imageGoogleAgent" : "imageOpenaiAgent",
             imageParams: { ...defaultImageParams, ...presentationStyle.imageParams },
         };
     },

package/lib/types/schema.d.ts

@@ -1146,6 +1146,7 @@ export declare const audioParamsSchema: z.ZodObject<{
     }>]>>;
     bgmVolume: z.ZodDefault<z.ZodNumber>;
     audioVolume: z.ZodDefault<z.ZodNumber>;
+    suppressSpeech: z.ZodDefault<z.ZodBoolean>;
 }, "strict", z.ZodTypeAny, {
     padding: number;
     introPadding: number;
@@ -1153,6 +1154,7 @@ export declare const audioParamsSchema: z.ZodObject<{
     outroPadding: number;
     bgmVolume: number;
     audioVolume: number;
+    suppressSpeech: boolean;
     bgm?: {
         url: string;
         kind: "url";
@@ -1186,6 +1188,7 @@ export declare const audioParamsSchema: z.ZodObject<{
     } | undefined;
     bgmVolume?: number | undefined;
     audioVolume?: number | undefined;
+    suppressSpeech?: boolean | undefined;
 }>;
 export declare const htmlPromptParamsSchema: z.ZodObject<{
     systemPrompt: z.ZodOptional<z.ZodDefault<z.ZodString>>;
@@ -2711,6 +2714,7 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
         }>]>>;
         bgmVolume: z.ZodDefault<z.ZodNumber>;
         audioVolume: z.ZodDefault<z.ZodNumber>;
+        suppressSpeech: z.ZodDefault<z.ZodBoolean>;
     }, "strict", z.ZodTypeAny, {
         padding: number;
         introPadding: number;
@@ -2718,6 +2722,7 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
         outroPadding: number;
         bgmVolume: number;
         audioVolume: number;
+        suppressSpeech: boolean;
         bgm?: {
             url: string;
             kind: "url";
@@ -2751,6 +2756,7 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
         } | undefined;
         bgmVolume?: number | undefined;
         audioVolume?: number | undefined;
+        suppressSpeech?: boolean | undefined;
     }>>;
 }, "strip", z.ZodTypeAny, {
     audioParams: {
@@ -2760,6 +2766,7 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
         outroPadding: number;
         bgmVolume: number;
         audioVolume: number;
+        suppressSpeech: boolean;
         bgm?: {
             url: string;
             kind: "url";
@@ -2885,6 +2892,7 @@ export declare const mulmoPresentationStyleSchema: z.ZodObject<{
         } | undefined;
         bgmVolume?: number | undefined;
         audioVolume?: number | undefined;
+        suppressSpeech?: boolean | undefined;
     } | undefined;
     movieParams?: {
         provider?: "openai" | "google" | "replicate" | undefined;
@@ -3250,6 +3258,7 @@ export declare const mulmoScriptSchema: z.ZodObject<{
         }>]>>;
         bgmVolume: z.ZodDefault<z.ZodNumber>;
         audioVolume: z.ZodDefault<z.ZodNumber>;
+        suppressSpeech: z.ZodDefault<z.ZodBoolean>;
     }, "strict", z.ZodTypeAny, {
         padding: number;
         introPadding: number;
@@ -3257,6 +3266,7 @@ export declare const mulmoScriptSchema: z.ZodObject<{
         outroPadding: number;
         bgmVolume: number;
         audioVolume: number;
+        suppressSpeech: boolean;
         bgm?: {
             url: string;
             kind: "url";
@@ -3290,6 +3300,7 @@ export declare const mulmoScriptSchema: z.ZodObject<{
         } | undefined;
         bgmVolume?: number | undefined;
         audioVolume?: number | undefined;
+        suppressSpeech?: boolean | undefined;
     }>>;
 } & {
     title: z.ZodOptional<z.ZodString>;
@@ -4390,6 +4401,7 @@ export declare const mulmoScriptSchema: z.ZodObject<{
         outroPadding: number;
         bgmVolume: number;
         audioVolume: number;
+        suppressSpeech: boolean;
         bgm?: {
             url: string;
             kind: "url";
@@ -4889,6 +4901,7 @@ export declare const mulmoScriptSchema: z.ZodObject<{
         } | undefined;
         bgmVolume?: number | undefined;
         audioVolume?: number | undefined;
+        suppressSpeech?: boolean | undefined;
     } | undefined;
     movieParams?: {
         provider?: "openai" | "google" | "replicate" | undefined;
@@ -5418,6 +5431,7 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             }>]>>;
             bgmVolume: z.ZodDefault<z.ZodNumber>;
             audioVolume: z.ZodDefault<z.ZodNumber>;
+            suppressSpeech: z.ZodDefault<z.ZodBoolean>;
         }, "strict", z.ZodTypeAny, {
             padding: number;
             introPadding: number;
@@ -5425,6 +5439,7 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -5458,6 +5473,7 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         }>>;
     } & {
         title: z.ZodOptional<z.ZodString>;
@@ -6558,6 +6574,7 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -7057,6 +7074,7 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         } | undefined;
         movieParams?: {
             provider?: "openai" | "google" | "replicate" | undefined;
@@ -7145,6 +7163,7 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -7655,6 +7674,7 @@ export declare const mulmoStudioSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         } | undefined;
         movieParams?: {
             provider?: "openai" | "google" | "replicate" | undefined;
@@ -8019,6 +8039,7 @@ export declare const mulmoScriptTemplateSchema: z.ZodObject<{
             }>]>>;
             bgmVolume: z.ZodDefault<z.ZodNumber>;
             audioVolume: z.ZodDefault<z.ZodNumber>;
+            suppressSpeech: z.ZodDefault<z.ZodBoolean>;
         }, "strict", z.ZodTypeAny, {
             padding: number;
             introPadding: number;
@@ -8026,6 +8047,7 @@ export declare const mulmoScriptTemplateSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -8059,6 +8081,7 @@ export declare const mulmoScriptTemplateSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         }>>;
     }, "strip", z.ZodTypeAny, {
         audioParams: {
@@ -8068,6 +8091,7 @@ export declare const mulmoScriptTemplateSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -8193,6 +8217,7 @@ export declare const mulmoScriptTemplateSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         } | undefined;
         movieParams?: {
             provider?: "openai" | "google" | "replicate" | undefined;
@@ -8246,6 +8271,7 @@ export declare const mulmoScriptTemplateSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -8377,6 +8403,7 @@ export declare const mulmoScriptTemplateSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         } | undefined;
         movieParams?: {
             provider?: "openai" | "google" | "replicate" | undefined;
@@ -8732,6 +8759,7 @@ export declare const mulmoScriptTemplateFileSchema: z.ZodObject<{
             }>]>>;
             bgmVolume: z.ZodDefault<z.ZodNumber>;
             audioVolume: z.ZodDefault<z.ZodNumber>;
+            suppressSpeech: z.ZodDefault<z.ZodBoolean>;
         }, "strict", z.ZodTypeAny, {
             padding: number;
             introPadding: number;
@@ -8739,6 +8767,7 @@ export declare const mulmoScriptTemplateFileSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -8772,6 +8801,7 @@ export declare const mulmoScriptTemplateFileSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         }>>;
     }, "strip", z.ZodTypeAny, {
         audioParams: {
@@ -8781,6 +8811,7 @@ export declare const mulmoScriptTemplateFileSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -8906,6 +8937,7 @@ export declare const mulmoScriptTemplateFileSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         } | undefined;
         movieParams?: {
             provider?: "openai" | "google" | "replicate" | undefined;
@@ -8962,6 +8994,7 @@ export declare const mulmoScriptTemplateFileSchema: z.ZodObject<{
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";
@@ -9094,6 +9127,7 @@ export declare const mulmoScriptTemplateFileSchema: z.ZodObject<{
             } | undefined;
             bgmVolume?: number | undefined;
             audioVolume?: number | undefined;
+            suppressSpeech?: boolean | undefined;
         } | undefined;
         movieParams?: {
             provider?: "openai" | "google" | "replicate" | undefined;

package/lib/types/schema.js

@@ -185,6 +185,7 @@ export const audioParamsSchema = z
     bgm: mediaSourceSchema.optional(),
     bgmVolume: z.number().default(0.2).describe("Volume of the background music"),
     audioVolume: z.number().default(1.0).describe("Volume of the audio"),
+    suppressSpeech: z.boolean().default(false).describe("Suppress speech generation"),
 })
     .strict();
 export const htmlPromptParamsSchema = z

package/lib/types/type.d.ts

@@ -49,7 +49,6 @@ export type MulmoStudioContext = {
     fileDirs: FileDirs;
     studio: MulmoStudio;
     lang?: string;
-    dryRun?: boolean;
     force: boolean;
     sessionState: MulmoSessionState;
     presentationStyle: MulmoPresentationStyle;
@@ -100,3 +99,27 @@ export type SessionProgressEvent = {
     inSession: boolean;
 };
 export type SessionProgressCallback = (change: SessionProgressEvent) => void;
+export interface FileObject {
+    baseDirPath: string;
+    mulmoFilePath: string;
+    mulmoFileDirPath: string;
+    outDirPath: string;
+    imageDirPath: string;
+    audioDirPath: string;
+    isHttpPath: boolean;
+    fileOrUrl: string;
+    outputStudioFilePath: string;
+    outputMultilingualFilePath: string;
+    presentationStylePath: string | undefined;
+    fileName: string;
+}
+export type InitOptions = {
+    b?: string;
+    o?: string;
+    i?: string;
+    a?: string;
+    file?: string;
+    l?: string;
+    c?: string;
+    p?: string;
+};

package/lib/utils/context.d.ts

@@ -0,0 +1,414 @@
+import type { MulmoScript, MulmoPresentationStyle, MulmoStudioMultiLingual } from "../types/type.js";
+import { FileObject } from "../types/index.js";
+export declare const fetchScript: (isHttpPath: boolean, mulmoFilePath: string, fileOrUrl: string) => Promise<MulmoScript | null>;
+export declare const getMultiLingual: (multilingualFilePath: string, studioBeatsLength: number) => MulmoStudioMultiLingual;
+export declare const getPresentationStyle: (presentationStylePath: string | undefined) => MulmoPresentationStyle | null;
+export declare const initializeContextFromFiles: (files: FileObject, raiseError: boolean, force?: boolean, caption?: string, lang?: string) => Promise<{
+    studio: {
+        beats: {
+            duration?: number | undefined;
+            hash?: string | undefined;
+            audioFile?: string | undefined;
+            imageFile?: string | undefined;
+            movieFile?: string | undefined;
+            captionFile?: string | undefined;
+        }[];
+        script: {
+            audioParams: {
+                padding: number;
+                introPadding: number;
+                closingPadding: number;
+                outroPadding: number;
+                bgmVolume: number;
+                audioVolume: number;
+                suppressSpeech: boolean;
+                bgm?: {
+                    url: string;
+                    kind: "url";
+                } | {
+                    kind: "base64";
+                    data: string;
+                } | {
+                    text: string;
+                    kind: "text";
+                } | {
+                    path: string;
+                    kind: "path";
+                } | undefined;
+            };
+            $mulmocast: {
+                version: "1.0";
+                credit?: "closing" | undefined;
+            };
+            canvasSize: {
+                width: number;
+                height: number;
+            };
+            speechParams: {
+                provider: "openai" | "nijivoice" | "google" | "elevenlabs";
+                speakers: Record<string, {
+                    voiceId: string;
+                    displayName?: Record<string, string> | undefined;
+                    speechOptions?: {
+                        speed?: number | undefined;
+                        instruction?: string | undefined;
+                    } | undefined;
+                    provider?: "openai" | "nijivoice" | "google" | "elevenlabs" | undefined;
+                }>;
+            };
+            beats: {
+                text: string;
+                speaker: string;
+                duration?: number | undefined;
+                speechOptions?: {
+                    speed?: number | undefined;
+                    instruction?: string | undefined;
+                } | undefined;
+                image?: {
+                    type: "markdown";
+                    markdown: string | string[];
+                } | {
+                    type: "web";
+                    url: string;
+                } | {
+                    type: "pdf";
+                    source: {
+                        url: string;
+                        kind: "url";
+                    } | {
+                        kind: "base64";
+                        data: string;
+                    } | {
+                        text: string;
+                        kind: "text";
+                    } | {
+                        path: string;
+                        kind: "path";
+                    };
+                } | {
+                    type: "image";
+                    source: {
+                        url: string;
+                        kind: "url";
+                    } | {
+                        kind: "base64";
+                        data: string;
+                    } | {
+                        text: string;
+                        kind: "text";
+                    } | {
+                        path: string;
+                        kind: "path";
+                    };
+                } | {
+                    type: "svg";
+                    source: {
+                        url: string;
+                        kind: "url";
+                    } | {
+                        kind: "base64";
+                        data: string;
+                    } | {
+                        text: string;
+                        kind: "text";
+                    } | {
+                        path: string;
+                        kind: "path";
+                    };
+                } | {
+                    type: "textSlide";
+                    slide: {
+                        title: string;
+                        subtitle?: string | undefined;
+                        bullets?: string[] | undefined;
+                    };
+                } | {
+                    type: "chart";
+                    title: string;
+                    chartData: Record<string, any>;
+                } | {
+                    code: {
+                        url: string;
+                        kind: "url";
+                    } | {
+                        kind: "base64";
+                        data: string;
+                    } | {
+                        text: string;
+                        kind: "text";
+                    } | {
+                        path: string;
+                        kind: "path";
+                    };
+                    type: "mermaid";
+                    title: string;
+                    appendix?: string[] | undefined;
+                } | {
+                    type: "html_tailwind";
+                    html: string | string[];
+                } | {
+                    type: "beat";
+                    id?: string | undefined;
+                } | {
+                    type: "movie";
+                    source: {
+                        url: string;
+                        kind: "url";
+                    } | {
+                        kind: "base64";
+                        data: string;
+                    } | {
+                        text: string;
+                        kind: "text";
+                    } | {
+                        path: string;
+                        kind: "path";
+                    };
+                    mixAudio: number;
+                } | undefined;
+                id?: string | undefined;
+                audio?: {
+                    type: "audio";
+                    source: {
+                        url: string;
+                        kind: "url";
+                    } | {
+                        kind: "base64";
+                        data: string;
+                    } | {
+                        text: string;
+                        kind: "text";
+                    } | {
+                        path: string;
+                        kind: "path";
+                    };
+                } | {
+                    type: "midi";
+                    source: string;
+                } | undefined;
+                description?: string | undefined;
+                imageParams?: {
+                    style?: string | undefined;
+                    model?: string | undefined;
+                    moderation?: string | undefined;
+                    images?: Record<string, {
+                        type: "image";
+                        source: {
+                            url: string;
+                            kind: "url";
+                        } | {
+                            kind: "base64";
+                            data: string;
+                        } | {
+                            text: string;
+                            kind: "text";
+                        } | {
+                            path: string;
+                            kind: "path";
+                        };
+                    }> | undefined;
+                } | undefined;
+                audioParams?: {
+                    padding?: number | undefined;
+                } | undefined;
+                movieParams?: {
+                    fillOption: {
+                        style: "aspectFit" | "aspectFill";
+                    };
+                } | undefined;
+                htmlImageParams?: {
+                    model?: string | undefined;
+                } | undefined;
+                textSlideParams?: {
+                    cssStyles: string | string[];
+                } | undefined;
+                captionParams?: {
+                    styles: string[];
+                    lang?: string | undefined;
+                } | undefined;
+                imageNames?: string[] | undefined;
+                imagePrompt?: string | undefined;
+                moviePrompt?: string | undefined;
+                htmlPrompt?: {
+                    prompt: string;
+                    data?: any;
+                    images?: Record<string, any> | undefined;
+                    systemPrompt?: string | undefined;
+                } | undefined;
+            }[];
+            lang?: string | undefined;
+            title?: string | undefined;
+            description?: string | undefined;
+            imageParams?: {
+                provider: "openai" | "google";
+                style?: string | undefined;
+                model?: string | undefined;
+                moderation?: string | undefined;
+                images?: Record<string, {
+                    type: "image";
+                    source: {
+                        url: string;
+                        kind: "url";
+                    } | {
+                        kind: "base64";
+                        data: string;
+                    } | {
+                        text: string;
+                        kind: "text";
+                    } | {
+                        path: string;
+                        kind: "path";
+                    };
+                }> | undefined;
+            } | undefined;
+            movieParams?: {
+                provider?: "openai" | "google" | "replicate" | undefined;
+                model?: string | undefined;
+                fillOption?: {
+                    style: "aspectFit" | "aspectFill";
+                } | undefined;
+                transition?: {
+                    type: "fade" | "slideout_left";
+                    duration: number;
+                } | undefined;
+            } | undefined;
+            htmlImageParams?: {
+                provider: "openai" | "anthropic";
+                model?: string | undefined;
+            } | undefined;
+            textSlideParams?: {
+                cssStyles: string | string[];
+            } | undefined;
+            captionParams?: {
+                styles: string[];
+                lang?: string | undefined;
+            } | undefined;
+            references?: {
+                type: "image" | "audio" | "article" | "paper" | "video";
+                url: string;
+                title?: string | undefined;
+                description?: string | undefined;
+            }[] | undefined;
+            imagePath?: string | undefined;
+            __test_invalid__?: boolean | undefined;
+        };
+        filename: string;
+    };
+    fileDirs: FileObject;
+    force: boolean;
+    lang: string | undefined;
+    sessionState: {
+        inSession: {
+            audio: boolean;
+            image: boolean;
+            video: boolean;
+            multiLingual: boolean;
+            caption: boolean;
+            pdf: boolean;
+        };
+        inBeatSession: {
+            audio: {};
+            image: {};
+            movie: {};
+            multiLingual: {};
+            caption: {};
+        };
+    };
+    presentationStyle: {
+        audioParams: {
+            padding: number;
+            introPadding: number;
+            closingPadding: number;
+            outroPadding: number;
+            bgmVolume: number;
+            audioVolume: number;
+            suppressSpeech: boolean;
+            bgm?: {
+                url: string;
+                kind: "url";
+            } | {
+                kind: "base64";
+                data: string;
+            } | {
+                text: string;
+                kind: "text";
+            } | {
+                path: string;
+                kind: "path";
+            } | undefined;
+        };
+        $mulmocast: {
+            version: "1.0";
+            credit?: "closing" | undefined;
+        };
+        canvasSize: {
+            width: number;
+            height: number;
+        };
+        speechParams: {
+            provider: "openai" | "nijivoice" | "google" | "elevenlabs";
+            speakers: Record<string, {
+                voiceId: string;
+                displayName?: Record<string, string> | undefined;
+                speechOptions?: {
+                    speed?: number | undefined;
+                    instruction?: string | undefined;
+                } | undefined;
+                provider?: "openai" | "nijivoice" | "google" | "elevenlabs" | undefined;
+            }>;
+        };
+        imageParams?: {
+            provider: "openai" | "google";
+            style?: string | undefined;
+            model?: string | undefined;
+            moderation?: string | undefined;
+            images?: Record<string, {
+                type: "image";
+                source: {
+                    url: string;
+                    kind: "url";
+                } | {
+                    kind: "base64";
+                    data: string;
+                } | {
+                    text: string;
+                    kind: "text";
+                } | {
+                    path: string;
+                    kind: "path";
+                };
+            }> | undefined;
+        } | undefined;
+        movieParams?: {
+            provider?: "openai" | "google" | "replicate" | undefined;
+            model?: string | undefined;
+            fillOption?: {
+                style: "aspectFit" | "aspectFill";
+            } | undefined;
+            transition?: {
+                type: "fade" | "slideout_left";
+                duration: number;
+            } | undefined;
+        } | undefined;
+        htmlImageParams?: {
+            provider: "openai" | "anthropic";
+            model?: string | undefined;
+        } | undefined;
+        textSlideParams?: {
+            cssStyles: string | string[];
+        } | undefined;
+        captionParams?: {
+            styles: string[];
+            lang?: string | undefined;
+        } | undefined;
+    };
+    multiLingual: {
+        multiLingualTexts: Record<string, {
+            text: string;
+            lang: string;
+            texts?: string[] | undefined;
+            ttsTexts?: string[] | undefined;
+            duration?: number | undefined;
+        }>;
+    }[];
+} | null>;

package/lib/utils/context.js

@@ -0,0 +1,96 @@
+import { GraphAILogger } from "graphai";
+import fs from "fs";
+import { readMulmoScriptFile, fetchMulmoScriptFile } from "./file.js";
+import { createOrUpdateStudioData } from "./preprocess.js";
+import { mulmoPresentationStyleSchema, mulmoStudioMultiLingualSchema } from "../types/index.js";
+export const fetchScript = async (isHttpPath, mulmoFilePath, fileOrUrl) => {
+    if (isHttpPath) {
+        const res = await fetchMulmoScriptFile(fileOrUrl);
+        if (!res.result || !res.script) {
+            GraphAILogger.info(`ERROR: HTTP error! ${res.status} ${fileOrUrl}`);
+            return null;
+        }
+        return res.script;
+    }
+    if (!fs.existsSync(mulmoFilePath)) {
+        GraphAILogger.info(`ERROR: File not exists ${mulmoFilePath}`);
+        return null;
+    }
+    return readMulmoScriptFile(mulmoFilePath, "ERROR: File does not exist " + mulmoFilePath)?.mulmoData ?? null;
+};
+export const getMultiLingual = (multilingualFilePath, studioBeatsLength) => {
+    if (fs.existsSync(multilingualFilePath)) {
+        const jsonData = readMulmoScriptFile(multilingualFilePath, "ERROR: File does not exist " + multilingualFilePath)?.mulmoData ?? null;
+        const dataSet = mulmoStudioMultiLingualSchema.parse(jsonData);
+        while (dataSet.length < studioBeatsLength) {
+            dataSet.push({ multiLingualTexts: {} });
+        }
+        dataSet.length = studioBeatsLength;
+        return dataSet;
+    }
+    return [...Array(studioBeatsLength)].map(() => ({ multiLingualTexts: {} }));
+};
+export const getPresentationStyle = (presentationStylePath) => {
+    if (presentationStylePath) {
+        if (!fs.existsSync(presentationStylePath)) {
+            throw new Error(`ERROR: File not exists ${presentationStylePath}`);
+        }
+        const jsonData = readMulmoScriptFile(presentationStylePath, "ERROR: File does not exist " + presentationStylePath)?.mulmoData ?? null;
+        return mulmoPresentationStyleSchema.parse(jsonData);
+    }
+    return null;
+};
+const initSessionState = () => {
+    return {
+        inSession: {
+            audio: false,
+            image: false,
+            video: false,
+            multiLingual: false,
+            caption: false,
+            pdf: false,
+        },
+        inBeatSession: {
+            audio: {},
+            image: {},
+            movie: {},
+            multiLingual: {},
+            caption: {},
+        },
+    };
+};
+const buildContext = (studio, files, presentationStyle, multiLingual, force, lang) => {
+    return {
+        studio,
+        fileDirs: files,
+        force: Boolean(force),
+        lang,
+        sessionState: initSessionState(),
+        presentationStyle: presentationStyle ?? studio.script,
+        multiLingual,
+    };
+};
+export const initializeContextFromFiles = async (files, raiseError, force, caption, lang) => {
+    const { fileName, isHttpPath, fileOrUrl, mulmoFilePath, outputStudioFilePath, presentationStylePath, outputMultilingualFilePath } = files;
+    // read mulmoScript, presentationStyle, currentStudio from files
+    const mulmoScript = await fetchScript(isHttpPath, mulmoFilePath, fileOrUrl);
+    if (!mulmoScript) {
+        return null;
+    }
+    const presentationStyle = getPresentationStyle(presentationStylePath);
+    // Create or update MulmoStudio file with MulmoScript
+    const currentStudio = readMulmoScriptFile(outputStudioFilePath);
+    try {
+        // validate mulmoStudioSchema. skip if __test_invalid__ is true
+        const studio = createOrUpdateStudioData(mulmoScript, currentStudio?.mulmoData, fileName, caption);
+        const multiLingual = getMultiLingual(outputMultilingualFilePath, studio.beats.length);
+        return buildContext(studio, files, presentationStyle, multiLingual, force, lang);
+    }
+    catch (error) {
+        GraphAILogger.info(`Error: invalid MulmoScript Schema: ${isHttpPath ? fileOrUrl : mulmoFilePath} \n ${error}`);
+        if (raiseError) {
+            throw error;
+        }
+        return null;
+    }
+};

package/lib/utils/ffmpeg_utils.d.ts

@@ -4,6 +4,7 @@ export type FfmpegContext = {
     inputCount: number;
     filterComplex: string[];
 };
+export declare const setFfmpegPath: (ffmpegPath: string) => void;
 export declare const FfmpegContextInit: () => FfmpegContext;
 export declare const FfmpegContextAddInput: (context: FfmpegContext, input: string, inputOptions?: string[]) => number;
 export declare const FfmpegContextPushFormattedAudio: (context: FfmpegContext, sourceId: string, outputId: string, duration?: number | undefined) => void;

package/lib/utils/ffmpeg_utils.js

@@ -1,5 +1,8 @@
 import ffmpeg from "fluent-ffmpeg";
 import { GraphAILogger } from "graphai";
+export const setFfmpegPath = (ffmpegPath) => {
+    ffmpeg.setFfmpegPath(ffmpegPath);
+};
 export const FfmpegContextInit = () => {
     return {
         command: ffmpeg(),

package/lib/utils/preprocess.d.ts

@@ -1,5 +1,5 @@
 import { MulmoStudio, MulmoScript } from "../types/index.js";
-export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, currentStudio: MulmoStudio | undefined, fileName: string) => {
+export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, currentStudio: MulmoStudio | undefined, fileName: string, videoCaption?: string) => {
     beats: {
         duration?: number | undefined;
         hash?: string | undefined;
@@ -16,6 +16,7 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
             outroPadding: number;
             bgmVolume: number;
             audioVolume: number;
+            suppressSpeech: boolean;
             bgm?: {
                 url: string;
                 kind: "url";

package/lib/utils/preprocess.js

@@ -1,5 +1,5 @@
 import { GraphAILogger } from "graphai";
-import { mulmoScriptSchema, mulmoStudioSchema } from "../types/index.js";
+import { mulmoScriptSchema, mulmoStudioSchema, mulmoCaptionParamsSchema } from "../types/index.js";
 const rebuildStudio = (currentStudio, mulmoScript, fileName) => {
     const isTest = process.env.NODE_ENV === "test";
     const parsed = isTest && currentStudio ? { data: mulmoStudioSchema.parse(currentStudio), success: true, error: null } : mulmoStudioSchema.safeParse(currentStudio);
@@ -36,7 +36,7 @@ const mulmoCredit = (speaker) => {
         },
     };
 };
-export const createOrUpdateStudioData = (_mulmoScript, currentStudio, fileName) => {
+export const createOrUpdateStudioData = (_mulmoScript, currentStudio, fileName, videoCaption) => {
     const mulmoScript = _mulmoScript.__test_invalid__ ? _mulmoScript : mulmoScriptSchema.parse(_mulmoScript); // validate and insert default value
     const studio = rebuildStudio(currentStudio, mulmoScript, fileName);
     // TODO: Move this code out of this function later
@@ -46,5 +46,11 @@ export const createOrUpdateStudioData = (_mulmoScript, currentStudio, fileName)
     }
     studio.script = mulmoScriptSchema.parse(mulmoScript); // update the script
     studio.beats = studio.script.beats.map((_, index) => studio.beats[index] ?? {});
+    if (videoCaption) {
+        studio.script.captionParams = mulmoCaptionParamsSchema.parse({
+            ...(studio.script.captionParams ?? {}),
+            lang: videoCaption,
+        });
+    }
     return studio;
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "0.0.26",
+  "version": "0.0.27",
   "description": "",
   "type": "module",
   "main": "lib/index.js",
@@ -78,12 +78,12 @@
     "google-auth-library": "^9.15.1",
     "graphai": "^2.0.9",
     "inquirer": "^12.6.3",
-    "marked": "^15.0.12",
+    "marked": "^16.0.0",
     "ora": "^8.2.0",
-    "puppeteer": "^24.11.0",
+    "puppeteer": "^24.11.1",
     "replicate": "^1.0.1",
     "yaml": "^2.8.0",
-    "yargs": "^17.7.2",
+    "yargs": "^18.0.0",
     "zod": "^3.25.67",
     "zod-to-json-schema": "^3.24.6"
   },