mulmocast 1.1.4 → 1.1.6

package/lib/types/schema.js

@@ -32,7 +32,9 @@ export const speakerDataSchema = z
     model: z.string().optional().describe("TTS model to use for this speaker"),
 })
     .strict();
-export const speakerDictionarySchema = z.record(speakerIdSchema, speakerDataSchema);
+export const speakerDictionarySchema = z.record(speakerIdSchema, speakerDataSchema.extend({
+    lang: z.record(langSchema, speakerDataSchema).optional(),
+}));
 export const mediaSourceSchema = z.discriminatedUnion("kind", [
     z.object({ kind: z.literal("url"), url: URLStringSchema }).strict(), // https://example.com/foo.pdf
     z.object({ kind: z.literal("base64"), data: z.string() }).strict(), // base64
@@ -172,6 +174,7 @@ export const mulmoOpenAIImageModelSchema = z
     .object({
     provider: z.literal("openai"),
     model: z.enum(provider2ImageAgent["openai"].models).optional(),
+    quality: z.enum(["low", "medium", "high", "auto"]).optional(),
 })
     .strict();
 // NOTE: This is for UI only. (until we figure out how to use it in mulmoImageParamsSchema)
@@ -185,6 +188,7 @@ export const mulmoImageParamsSchema = z
     .object({
     provider: text2ImageProviderSchema, // has default value
     model: z.string().optional(), // default: provider specific
+    quality: z.string().optional(), // optional image quality (model specific)
     style: z.string().optional(), // optional image style
     moderation: z.string().optional(), // optional image style
     images: mulmoImageParamsImagesSchema.optional(),
@@ -367,7 +371,7 @@ export const mulmoScriptSchema = mulmoPresentationStyleSchema
     title: z.string().optional(),
     description: z.string().optional(),
     references: z.array(mulmoReferenceSchema).optional(),
-    lang: langSchema.optional(), // default "en"
+    lang: langSchema, // required (default WAS "en")
     beats: z.array(mulmoBeatSchema).min(1),
     // TODO: Delete it later
     imagePath: z.string().optional(), // for keynote images movie ??

package/lib/types/type.d.ts

@@ -53,7 +53,7 @@ export type FileDirs = {
 export type MulmoStudioContext = {
     fileDirs: FileDirs;
     studio: MulmoStudio;
-    lang?: string;
+    lang: string;
     force: boolean;
     sessionState: MulmoSessionState;
     presentationStyle: MulmoPresentationStyle;

package/lib/utils/const.js

@@ -4,7 +4,7 @@ export const imageDirName = "images";
 export const cacheDirName = "cache";
 export const pdf_modes = ["slide", "talk", "handout"];
 export const pdf_sizes = ["letter", "a4"];
-export const languages = ["en", "ja"];
+export const languages = ["en", "ja", "fr", "es"];
 export const storyToScriptGenerateMode = {
     stepWise: "step_wise",
     oneStep: "one_step",

package/lib/utils/context.d.ts

@@ -1,9 +1,351 @@
 import type { MulmoScript, MulmoPresentationStyle, MulmoStudioMultiLingual } from "../types/type.js";
 import { FileObject } from "../types/index.js";
+export declare const createStudioData: (_mulmoScript: MulmoScript, fileName: string, videoCaptionLang?: string, presentationStyle?: MulmoPresentationStyle | null) => {
+    beats: {
+        duration?: number | undefined;
+        startAt?: number | undefined;
+        hash?: string | undefined;
+        audioDuration?: number | undefined;
+        movieDuration?: number | undefined;
+        silenceDuration?: number | undefined;
+        hasMovieAudio?: boolean | undefined;
+        audioFile?: string | undefined;
+        imageFile?: string | undefined;
+        movieFile?: string | undefined;
+        soundEffectFile?: string | undefined;
+        lipSyncFile?: string | undefined;
+        captionFile?: string | undefined;
+    }[];
+    script: {
+        lang: string;
+        imageParams: {
+            provider: string;
+            model?: string | undefined;
+            style?: string | undefined;
+            quality?: 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";
+                };
+            } | {
+                type: "imagePrompt";
+                prompt: string;
+            }> | undefined;
+        };
+        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;
+        };
+        movieParams: {
+            provider?: string | undefined;
+            model?: string | undefined;
+            fillOption?: {
+                style: "aspectFit" | "aspectFill";
+            } | undefined;
+            transition?: {
+                type: "fade" | "slideout_left";
+                duration: number;
+            } | undefined;
+        };
+        soundEffectParams: {
+            provider?: string | undefined;
+            model?: string | undefined;
+        };
+        $mulmocast: {
+            version: "1.1";
+            credit?: "closing" | undefined;
+        };
+        canvasSize: {
+            width: number;
+            height: number;
+        };
+        speechParams: {
+            speakers: Record<string, {
+                voiceId: string;
+                lang?: Record<string, {
+                    voiceId: string;
+                    displayName?: Record<string, string> | undefined;
+                    isDefault?: boolean | undefined;
+                    speechOptions?: {
+                        speed?: number | undefined;
+                        instruction?: string | undefined;
+                    } | undefined;
+                    provider?: string | undefined;
+                    model?: string | undefined;
+                }> | undefined;
+                displayName?: Record<string, string> | undefined;
+                isDefault?: boolean | undefined;
+                speechOptions?: {
+                    speed?: number | undefined;
+                    instruction?: string | undefined;
+                } | undefined;
+                provider?: string | undefined;
+                model?: string | undefined;
+            }>;
+        };
+        beats: {
+            text: string;
+            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: "movie";
+                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: "voice_over";
+                startAt?: number | undefined;
+            } | 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;
+            duration?: number | undefined;
+            speechOptions?: {
+                speed?: number | undefined;
+                instruction?: string | undefined;
+            } | undefined;
+            id?: string | undefined;
+            imagePrompt?: string | undefined;
+            speaker?: string | undefined;
+            description?: string | undefined;
+            imageParams?: {
+                provider: string;
+                model?: string | undefined;
+                style?: string | undefined;
+                quality?: 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";
+                    };
+                } | {
+                    type: "imagePrompt";
+                    prompt: string;
+                }> | undefined;
+            } | undefined;
+            audioParams?: {
+                movieVolume: number;
+                padding?: number | undefined;
+            } | undefined;
+            movieParams?: {
+                speed?: number | undefined;
+                provider?: string | undefined;
+                model?: string | undefined;
+                fillOption?: {
+                    style: "aspectFit" | "aspectFill";
+                } | undefined;
+            } | undefined;
+            soundEffectParams?: {
+                provider?: string | undefined;
+                model?: string | undefined;
+            } | undefined;
+            lipSyncParams?: {
+                provider?: string | undefined;
+                model?: string | undefined;
+            } | undefined;
+            htmlImageParams?: {
+                model?: string | undefined;
+            } | undefined;
+            textSlideParams?: {
+                cssStyles: string | string[];
+            } | undefined;
+            captionParams?: {
+                styles: string[];
+                lang?: string | undefined;
+            } | undefined;
+            imageNames?: string[] | undefined;
+            moviePrompt?: string | undefined;
+            soundEffectPrompt?: string | undefined;
+            htmlPrompt?: {
+                prompt: string;
+                data?: any;
+                images?: Record<string, any> | undefined;
+                systemPrompt?: string | undefined;
+            } | undefined;
+            enableLipSync?: boolean | undefined;
+        }[];
+        title?: string | undefined;
+        description?: string | undefined;
+        lipSyncParams?: {
+            provider?: string | undefined;
+            model?: string | undefined;
+        } | undefined;
+        htmlImageParams?: {
+            provider: string;
+            model?: string | undefined;
+        } | undefined;
+        textSlideParams?: {
+            cssStyles: string | string[];
+        } | undefined;
+        captionParams?: {
+            styles: string[];
+            lang?: string | undefined;
+        } | undefined;
+        references?: {
+            type: string;
+            url: string;
+            title?: string | undefined;
+            description?: string | undefined;
+        }[] | undefined;
+        imagePath?: string | undefined;
+        __test_invalid__?: boolean | undefined;
+    };
+    filename: string;
+};
 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<{
+export declare const initializeContextFromFiles: (files: FileObject, raiseError: boolean, force?: boolean, captionLang?: string, targetLang?: string) => Promise<{
     studio: {
         beats: {
             duration?: number | undefined;
@@ -21,10 +363,12 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
             captionFile?: string | undefined;
         }[];
         script: {
+            lang: string;
             imageParams: {
                 provider: string;
                 model?: string | undefined;
                 style?: string | undefined;
+                quality?: string | undefined;
                 moderation?: string | undefined;
                 images?: Record<string, {
                     type: "image";
@@ -94,6 +438,17 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
             speechParams: {
                 speakers: Record<string, {
                     voiceId: string;
+                    lang?: Record<string, {
+                        voiceId: string;
+                        displayName?: Record<string, string> | undefined;
+                        isDefault?: boolean | undefined;
+                        speechOptions?: {
+                            speed?: number | undefined;
+                            instruction?: string | undefined;
+                        } | undefined;
+                        provider?: string | undefined;
+                        model?: string | undefined;
+                    }> | undefined;
                     displayName?: Record<string, string> | undefined;
                     isDefault?: boolean | undefined;
                     speechOptions?: {
@@ -242,6 +597,7 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                     provider: string;
                     model?: string | undefined;
                     style?: string | undefined;
+                    quality?: string | undefined;
                     moderation?: string | undefined;
                     images?: Record<string, {
                         type: "image";
@@ -304,7 +660,6 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
                 } | undefined;
                 enableLipSync?: boolean | undefined;
             }[];
-            lang?: string | undefined;
             title?: string | undefined;
             description?: string | undefined;
             lipSyncParams?: {
@@ -333,35 +688,22 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
         };
         filename: string;
     };
+    multiLingual: {
+        multiLingualTexts: Record<string, {
+            text: string;
+            lang: string;
+            texts?: string[] | undefined;
+            ttsTexts?: string[] | undefined;
+            duration?: number | undefined;
+        }>;
+    }[];
     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: {};
-            html: {};
-            imageReference: {};
-            soundEffect: {};
-            lipSync: {};
-        };
-    };
     presentationStyle: {
         imageParams: {
             provider: string;
             model?: string | undefined;
             style?: string | undefined;
+            quality?: string | undefined;
             moderation?: string | undefined;
             images?: Record<string, {
                 type: "image";
@@ -431,6 +773,17 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
         speechParams: {
             speakers: Record<string, {
                 voiceId: string;
+                lang?: Record<string, {
+                    voiceId: string;
+                    displayName?: Record<string, string> | undefined;
+                    isDefault?: boolean | undefined;
+                    speechOptions?: {
+                        speed?: number | undefined;
+                        instruction?: string | undefined;
+                    } | undefined;
+                    provider?: string | undefined;
+                    model?: string | undefined;
+                }> | undefined;
                 displayName?: Record<string, string> | undefined;
                 isDefault?: boolean | undefined;
                 speechOptions?: {
@@ -457,13 +810,27 @@ export declare const initializeContextFromFiles: (files: FileObject, raiseError:
             lang?: string | undefined;
         } | undefined;
     };
-    multiLingual: {
-        multiLingualTexts: Record<string, {
-            text: string;
-            lang: string;
-            texts?: string[] | undefined;
-            ttsTexts?: string[] | undefined;
-            duration?: number | undefined;
-        }>;
-    }[];
+    sessionState: {
+        inSession: {
+            audio: boolean;
+            image: boolean;
+            video: boolean;
+            multiLingual: boolean;
+            caption: boolean;
+            pdf: boolean;
+        };
+        inBeatSession: {
+            audio: {};
+            image: {};
+            movie: {};
+            multiLingual: {};
+            caption: {};
+            html: {};
+            imageReference: {};
+            soundEffect: {};
+            lipSync: {};
+        };
+    };
+    force: boolean;
+    lang: string;
 } | null>;