mulmocast 1.1.6 → 1.1.7

package/lib/types/schema.js

@@ -1,5 +1,6 @@
 import { z } from "zod";
 import { htmlLLMProvider, provider2TTSAgent, provider2ImageAgent, provider2MovieAgent, defaultProviders, provider2SoundEffectAgent, } from "../utils/provider2agent.js";
+import { currentMulmoScriptVersion } from "../utils/const.js";
 export const langSchema = z.string();
 const URLStringSchema = z.string().url();
 export const localizedTextSchema = z
@@ -8,7 +9,8 @@ export const localizedTextSchema = z
     lang: langSchema,
     // caption: z.string(),
     texts: z.array(z.string()).optional(),
-    ttsTexts: z.array(z.string()).optional(),
+    ttsTexts: z.array(z.string()).optional(), // TODO remove
+    cacheKey: z.string(),
     duration: z.number().optional(), // generated // video duration time(ms)
     // filename: z.string().optional(), // generated //
 })
@@ -283,7 +285,7 @@ export const mulmoCanvasDimensionSchema = z
 // export const voiceMapSchema = z.record(speakerIdSchema, z.string())
 export const mulmoCastCreditSchema = z
     .object({
-    version: z.literal("1.1"),
+    version: z.literal(currentMulmoScriptVersion),
     credit: z.literal("closing").optional(),
 })
     .strict();
@@ -398,8 +400,13 @@ export const mulmoStudioBeatSchema = z
     .strict();
 export const mulmoStudioMultiLingualDataSchema = z.object({
     multiLingualTexts: multiLingualTextsSchema,
+    cacheKey: z.string().optional(),
 });
 export const mulmoStudioMultiLingualSchema = z.array(mulmoStudioMultiLingualDataSchema).min(1);
+export const mulmoStudioMultiLingualFileSchema = z.object({
+    version: z.literal(currentMulmoScriptVersion),
+    multiLingual: mulmoStudioMultiLingualSchema,
+});
 export const mulmoSessionStateSchema = z.object({
     inSession: z.object({
         audio: z.boolean(),

package/lib/types/type.d.ts

@@ -1,4 +1,4 @@
-import { langSchema, localizedTextSchema, mulmoBeatSchema, mulmoScriptSchema, mulmoStudioSchema, mulmoStudioBeatSchema, mulmoStoryboardSchema, mulmoStoryboardSceneSchema, mulmoStudioMultiLingualSchema, mulmoStudioMultiLingualDataSchema, speakerDictionarySchema, mulmoImageParamsSchema, mulmoImageParamsImagesSchema, mulmoFillOptionSchema, mulmoMovieParamsSchema, textSlideParamsSchema, speechOptionsSchema, speakerDataSchema, mulmoCanvasDimensionSchema, mulmoPromptTemplateSchema, mulmoPromptTemplateFileSchema, text2ImageProviderSchema, text2HtmlImageProviderSchema, text2MovieProviderSchema, text2SpeechProviderSchema, mulmoPresentationStyleSchema, multiLingualTextsSchema, mulmoImageAssetSchema, mulmoMermaidMediaSchema, mulmoTextSlideMediaSchema, mulmoMarkdownMediaSchema, mulmoImageMediaSchema, mulmoChartMediaSchema, mediaSourceSchema, mulmoSessionStateSchema, mulmoOpenAIImageModelSchema, mulmoGoogleImageModelSchema, mulmoGoogleMovieModelSchema, mulmoReplicateMovieModelSchema, mulmoImagePromptMediaSchema } from "./schema.js";
+import { langSchema, localizedTextSchema, mulmoBeatSchema, mulmoScriptSchema, mulmoStudioSchema, mulmoStudioBeatSchema, mulmoStoryboardSchema, mulmoStoryboardSceneSchema, mulmoStudioMultiLingualSchema, mulmoStudioMultiLingualDataSchema, mulmoStudioMultiLingualFileSchema, speakerDictionarySchema, mulmoImageParamsSchema, mulmoImageParamsImagesSchema, mulmoFillOptionSchema, mulmoMovieParamsSchema, textSlideParamsSchema, speechOptionsSchema, speakerDataSchema, mulmoCanvasDimensionSchema, mulmoPromptTemplateSchema, mulmoPromptTemplateFileSchema, text2ImageProviderSchema, text2HtmlImageProviderSchema, text2MovieProviderSchema, text2SpeechProviderSchema, mulmoPresentationStyleSchema, multiLingualTextsSchema, mulmoImageAssetSchema, mulmoMermaidMediaSchema, mulmoTextSlideMediaSchema, mulmoMarkdownMediaSchema, mulmoImageMediaSchema, mulmoChartMediaSchema, mediaSourceSchema, 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";
@@ -28,6 +28,7 @@ export type MulmoPromptTemplate = z.infer<typeof mulmoPromptTemplateSchema>;
 export type MulmoPromptTemplateFile = z.infer<typeof mulmoPromptTemplateFileSchema>;
 export type MulmoStudioMultiLingual = z.infer<typeof mulmoStudioMultiLingualSchema>;
 export type MulmoStudioMultiLingualData = z.infer<typeof mulmoStudioMultiLingualDataSchema>;
+export type MulmoStudioMultiLingualFile = z.infer<typeof mulmoStudioMultiLingualFileSchema>;
 export type MultiLingualTexts = z.infer<typeof multiLingualTextsSchema>;
 export type MulmoMovieParams = z.infer<typeof mulmoMovieParamsSchema>;
 export type MulmoOpenAIImageModel = z.infer<typeof mulmoOpenAIImageModelSchema>;

package/lib/utils/const.d.ts

@@ -1,3 +1,4 @@
+export declare const currentMulmoScriptVersion = "1.1";
 export declare const outDirName = "output";
 export declare const audioDirName = "audio";
 export declare const imageDirName = "images";

package/lib/utils/const.js

@@ -1,3 +1,4 @@
+export const currentMulmoScriptVersion = "1.1";
 export const outDirName = "output";
 export const audioDirName = "audio";
 export const imageDirName = "images";

package/lib/utils/context.d.ts

@@ -1,5 +1,4 @@
-import type { MulmoScript, MulmoPresentationStyle, MulmoStudioMultiLingual } from "../types/type.js";
-import { FileObject } from "../types/index.js";
+import type { MulmoScript, MulmoPresentationStyle, MulmoStudioMultiLingual, FileObject } from "../types/type.js";
 export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: string, videoCaptionLang?: string, presentationStyle?: MulmoPresentationStyle | null) => {
     beats: {
         duration?: number | undefined;
@@ -20,10 +19,6 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
         lang: string;
         imageParams: {
             provider: string;
-            model?: string | undefined;
-            style?: string | undefined;
-            quality?: string | undefined;
-            moderation?: string | undefined;
             images?: Record<string, {
                 type: "image";
                 source: {
@@ -43,6 +38,10 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                 type: "imagePrompt";
                 prompt: string;
             }> | undefined;
+            model?: string | undefined;
+            style?: string | undefined;
+            quality?: string | undefined;
+            moderation?: string | undefined;
         };
         audioParams: {
             padding: number;
@@ -182,12 +181,12 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                     kind: "path";
                 };
             } | {
-                type: "textSlide";
                 slide: {
                     title: string;
                     subtitle?: string | undefined;
                     bullets?: string[] | undefined;
                 };
+                type: "textSlide";
             } | {
                 type: "chart";
                 title: string;
@@ -249,10 +248,6 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
             description?: string | undefined;
             imageParams?: {
                 provider: string;
-                model?: string | undefined;
-                style?: string | undefined;
-                quality?: string | undefined;
-                moderation?: string | undefined;
                 images?: Record<string, {
                     type: "image";
                     source: {
@@ -272,6 +267,10 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
                     type: "imagePrompt";
                     prompt: string;
                 }> | undefined;
+                model?: string | undefined;
+                style?: string | undefined;
+                quality?: string | undefined;
+                moderation?: string | undefined;
             } | undefined;
             audioParams?: {
                 movieVolume: number;
@@ -308,8 +307,8 @@ export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: str
             soundEffectPrompt?: string | undefined;
             htmlPrompt?: {
                 prompt: string;
-                data?: any;
                 images?: Record<string, any> | undefined;
+                data?: any;
                 systemPrompt?: string | undefined;
             } | undefined;
             enableLipSync?: boolean | undefined;
@@ -366,10 +365,6 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
             lang: string;
             imageParams: {
                 provider: string;
-                model?: string | undefined;
-                style?: string | undefined;
-                quality?: string | undefined;
-                moderation?: string | undefined;
                 images?: Record<string, {
                     type: "image";
                     source: {
@@ -389,6 +384,10 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     type: "imagePrompt";
                     prompt: string;
                 }> | undefined;
+                model?: string | undefined;
+                style?: string | undefined;
+                quality?: string | undefined;
+                moderation?: string | undefined;
             };
             audioParams: {
                 padding: number;
@@ -528,12 +527,12 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                         kind: "path";
                     };
                 } | {
-                    type: "textSlide";
                     slide: {
                         title: string;
                         subtitle?: string | undefined;
                         bullets?: string[] | undefined;
                     };
+                    type: "textSlide";
                 } | {
                     type: "chart";
                     title: string;
@@ -595,10 +594,6 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 description?: string | undefined;
                 imageParams?: {
                     provider: string;
-                    model?: string | undefined;
-                    style?: string | undefined;
-                    quality?: string | undefined;
-                    moderation?: string | undefined;
                     images?: Record<string, {
                         type: "image";
                         source: {
@@ -618,6 +613,10 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                         type: "imagePrompt";
                         prompt: string;
                     }> | undefined;
+                    model?: string | undefined;
+                    style?: string | undefined;
+                    quality?: string | undefined;
+                    moderation?: string | undefined;
                 } | undefined;
                 audioParams?: {
                     movieVolume: number;
@@ -654,8 +653,8 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 soundEffectPrompt?: string | undefined;
                 htmlPrompt?: {
                     prompt: string;
-                    data?: any;
                     images?: Record<string, any> | undefined;
+                    data?: any;
                     systemPrompt?: string | undefined;
                 } | undefined;
                 enableLipSync?: boolean | undefined;
@@ -692,19 +691,17 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
         multiLingualTexts: Record<string, {
             text: string;
             lang: string;
+            cacheKey: string;
             texts?: string[] | undefined;
             ttsTexts?: string[] | undefined;
             duration?: number | undefined;
         }>;
+        cacheKey?: string | undefined;
     }[];
     fileDirs: FileObject;
     presentationStyle: {
         imageParams: {
             provider: string;
-            model?: string | undefined;
-            style?: string | undefined;
-            quality?: string | undefined;
-            moderation?: string | undefined;
             images?: Record<string, {
                 type: "image";
                 source: {
@@ -724,6 +721,10 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 type: "imagePrompt";
                 prompt: string;
             }> | undefined;
+            model?: string | undefined;
+            style?: string | undefined;
+            quality?: string | undefined;
+            moderation?: string | undefined;
         };
         audioParams: {
             padding: number;

package/lib/utils/context.js

@@ -1,9 +1,8 @@
 import { GraphAILogger } from "graphai";
 import fs from "fs";
 import { readMulmoScriptFile, fetchMulmoScriptFile } from "./file.js";
-import { mulmoStudioSchema, mulmoCaptionParamsSchema } from "../types/index.js";
-import { MulmoPresentationStyleMethods, MulmoScriptMethods } from "../methods/index.js";
-import { mulmoPresentationStyleSchema, mulmoStudioMultiLingualSchema } from "../types/index.js";
+import { mulmoStudioSchema, mulmoCaptionParamsSchema, mulmoPresentationStyleSchema } from "../types/schema.js";
+import { MulmoPresentationStyleMethods, MulmoScriptMethods, MulmoStudioMultiLingualMethod } from "../methods/index.js";
 const mulmoCredit = (speaker) => {
     return {
         speaker,
@@ -92,12 +91,7 @@ export const getMultiLingual = (multilingualFilePath, studioBeatsLength) => {
         return [...Array(studioBeatsLength)].map(() => ({ multiLingualTexts: {} }));
     }
     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 MulmoStudioMultiLingualMethod.validate(jsonData, studioBeatsLength);
 };
 export const getPresentationStyle = (presentationStylePath) => {
     if (!presentationStylePath) {

package/lib/utils/image_plugins/mermaid.js

@@ -5,7 +5,7 @@ import { parrotingImagePath } from "./utils.js";
 export const imageType = "mermaid";
 const processMermaid = async (params) => {
     const { beat, imagePath, canvasSize, context, textSlideStyle } = params;
-    if (!beat || !beat.image || beat.image.type !== imageType)
+    if (!beat?.image || beat.image.type !== imageType)
         return;
     const template = getHTMLFile("mermaid");
     const diagram_code = await MulmoMediaSourceMethods.getText(beat.image.code, context);

package/lib/utils/image_plugins/source.js

@@ -3,7 +3,7 @@ import { MulmoMediaSourceMethods } from "../../methods/mulmo_media_source.js";
 export const processSource = (imageType) => {
     return (params) => {
         const { beat, context } = params;
-        if (!beat || !beat.image || beat.image.type !== imageType)
+        if (!beat?.image || beat.image.type !== imageType)
             return;
         const path = MulmoMediaSourceMethods.resolve(beat.image.source, context);
         if (path) {

package/lib/utils/string.js

@@ -54,4 +54,5 @@ export const replacementsJa = [
     { from: "危険な面", to: "危険なめん" },
     { from: "その通り！", to: "その通り。" },
     { from: "%", to: "パーセント" },
+    { from: "IPO", to: "アイピーオー" },
 ];

package/lib/utils/utils.js

@@ -21,12 +21,8 @@ export const text2hash = (input) => {
     return crypto.createHash("sha256").update(input).digest("hex");
 };
 export const localizedText = (beat, multiLingualData, lang) => {
-    if (lang &&
-        multiLingualData &&
-        multiLingualData?.multiLingualTexts &&
-        multiLingualData?.multiLingualTexts[lang] &&
-        multiLingualData?.multiLingualTexts[lang].text) {
-        return multiLingualData?.multiLingualTexts[lang].text;
+    if (lang && multiLingualData?.multiLingualTexts?.[lang]?.text) {
+        return multiLingualData.multiLingualTexts[lang].text;
     }
     return beat.text;
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "1.1.6",
+  "version": "1.1.7",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -105,7 +105,7 @@
     "prettier": "^3.6.2",
     "ts-node": "^10.9.2",
     "tsx": "^4.20.3",
-    "typescript": "^5.7.3",
+    "typescript": "^5.9.2",
     "typescript-eslint": "^8.37.0"
   },
   "engines": {

package/scripts/test/test_lang.json

@@ -66,7 +66,7 @@
       }
     },
     {
-      "text": "The yield of TSMC's 3nm process exceeeds 70%.",
+      "text": "The yield of TSMC's 3nm process exceeds 70%.",
       "image": {
         "type": "textSlide",
         "slide": {