mulmocast 2.1.15 → 2.1.16

package/lib/actions/audio.js

@@ -8,7 +8,7 @@ import { fileCacheAgentFilter, nijovoiceTextAgentFilter } from "../utils/filters
 import { getAudioArtifactFilePath, getAudioFilePath, getOutputStudioFilePath, resolveDirPath, defaultBGMPath, mkdir, writingMessage } from "../utils/file.js";
 import { localizedText, settings2GraphAIConfig } from "../utils/utils.js";
 import { text2hash } from "../utils/utils_node.js";
-import { provider2TTSAgent } from "../utils/provider2agent.js";
+import { provider2TTSAgent } from "../types/provider2agent.js";
 import { invalidAudioSourceError } from "../utils/error_cause.js";
 import { MulmoStudioContextMethods } from "../methods/mulmo_studio_context.js";
 import { MulmoMediaSourceMethods } from "../methods/mulmo_media_source.js";

package/lib/actions/bundle.d.ts

@@ -1,2 +1,5 @@
 import { type MulmoStudioContext } from "../types/index.js";
-export declare const mulmoViewerBundle: (context: MulmoStudioContext) => Promise<void>;
+export type MulmoViewerBundleOptions = {
+    skipZip?: boolean;
+};
+export declare const mulmoViewerBundle: (context: MulmoStudioContext, options?: MulmoViewerBundleOptions) => Promise<void>;

package/lib/actions/bundle.js

@@ -4,7 +4,7 @@ import { GraphAILogger } from "graphai";
 import { listLocalizedAudioPaths } from "./audio.js";
 import { mkdir } from "../utils/file.js";
 import { ZipBuilder } from "../utils/zip.js";
-import { bundleTargetLang } from "../utils/const.js";
+import { bundleTargetLang } from "../types/const.js";
 import { createSilentAudio } from "../utils/ffmpeg_utils.js";
 import { silentMp3 } from "../utils/context.js";
 const downloadFile = async (url, destPath) => {
@@ -27,9 +27,12 @@ const processBgm = async (bgm, outDir, baseDir, zipper) => {
             return undefined;
         }
         const fileName = path.basename(bgm.path);
-        const destPath = path.resolve(outDir, fileName);
-        fs.copyFileSync(sourcePath, destPath);
-        zipper.addFile(sourcePath, fileName);
+        if (zipper) {
+            zipper.addFile(sourcePath, fileName);
+        }
+        else {
+            fs.copyFileSync(sourcePath, path.resolve(outDir, fileName));
+        }
         return fileName;
     }
     else if (bgm.kind === "url") {
@@ -37,7 +40,7 @@ const processBgm = async (bgm, outDir, baseDir, zipper) => {
         const fileName = path.basename(new URL(bgm.url).pathname) || "bgm.mp3";
         const destPath = path.resolve(outDir, fileName);
         await downloadFile(bgm.url, destPath);
-        zipper.addFile(destPath);
+        zipper?.addFile(destPath);
         return fileName;
     }
     // base64 or other formats are not supported
@@ -52,12 +55,16 @@ const imageSourceMappings = [
     ["lipSyncFile", "videoWithAudioSource"],
     ["htmlImageFile", "htmlImageSource"],
 ];
-export const mulmoViewerBundle = async (context) => {
-    const isZip = true;
+export const mulmoViewerBundle = async (context, options = {}) => {
+    const { skipZip = false } = options;
     const outDir = context.fileDirs.outDirPath;
     const baseDir = context.fileDirs.baseDirPath;
+    const filename = context.studio.filename;
     mkdir(outDir);
-    const zipper = new ZipBuilder(path.resolve(outDir, zipFileName));
+    // Bundle directory: output/<script_name>/
+    const bundleDir = path.resolve(outDir, filename);
+    mkdir(bundleDir);
+    const zipper = skipZip ? undefined : new ZipBuilder(path.resolve(bundleDir, zipFileName));
     // text
     const resultJson = [];
     context.studio.script.beats.forEach((beat, index) => {
@@ -77,13 +84,17 @@ export const mulmoViewerBundle = async (context) => {
                 }
                 if (fileName === "silent300.mp3") {
                     // Download from GitHub URL
-                    const destPath = path.resolve(outDir, fileName);
+                    const destPath = path.resolve(bundleDir, fileName);
                     await downloadFile(silentMp3, destPath);
-                    zipper.addFile(destPath, fileName);
+                    zipper?.addFile(destPath, fileName);
                 }
                 else if (fs.existsSync(audio)) {
-                    fs.copyFileSync(audio, path.resolve(outDir, fileName));
-                    zipper.addFile(audio, fileName);
+                    if (zipper) {
+                        zipper.addFile(audio, fileName);
+                    }
+                    else {
+                        fs.copyFileSync(audio, path.resolve(bundleDir, fileName));
+                    }
                 }
             }
         }));
@@ -96,13 +107,17 @@ export const mulmoViewerBundle = async (context) => {
             if (typeof value === "string") {
                 data[source] = path.basename(value);
                 if (fs.existsSync(value)) {
-                    fs.copyFileSync(value, path.resolve(outDir, path.basename(value)));
-                    zipper.addFile(value);
+                    if (zipper) {
+                        zipper.addFile(value);
+                    }
+                    else {
+                        fs.copyFileSync(value, path.resolve(bundleDir, path.basename(value)));
+                    }
                 }
             }
         });
     });
-    // silent
+    // silent - generated files always go to bundleDir
     await Promise.all(context.studio.script.beats.map(async (__, index) => {
         const data = resultJson[index];
         if (data.audioSources &&
@@ -111,9 +126,9 @@ export const mulmoViewerBundle = async (context) => {
             data.videoWithAudioSource === undefined &&
             data.duration) {
             const file = `silent_${index}.mp3`;
-            const audioFile = path.resolve(outDir, file);
+            const audioFile = path.resolve(bundleDir, file);
             await createSilentAudio(audioFile, data.duration);
-            zipper.addFile(audioFile);
+            zipper?.addFile(audioFile);
             data.audioSources.ja = file;
             data.audioSources.en = file;
         }
@@ -127,11 +142,12 @@ export const mulmoViewerBundle = async (context) => {
         });
     });
     // BGM
-    const bgmFileName = await processBgm(context.studio?.script.audioParams?.bgm, outDir, baseDir, zipper);
+    const bgmFileName = await processBgm(context.studio?.script.audioParams?.bgm, bundleDir, baseDir, zipper);
     const bundleData = { beats: resultJson, bgmSource: bgmFileName, title: context.studio.script.title };
-    fs.writeFileSync(path.resolve(outDir, viewJsonFileName), JSON.stringify(bundleData, null, 2));
-    zipper.addFile(path.resolve(outDir, viewJsonFileName));
-    if (isZip) {
+    const viewJsonPath = path.resolve(bundleDir, viewJsonFileName);
+    fs.writeFileSync(viewJsonPath, JSON.stringify(bundleData, null, 2));
+    zipper?.addFile(viewJsonPath);
+    if (zipper) {
         await zipper.finalize();
     }
 };

package/lib/actions/translate.js

@@ -7,7 +7,7 @@ import { fileWriteAgent } from "@graphai/vanilla_node_agents";
 import { splitText } from "../utils/string.js";
 import { settings2GraphAIConfig, beatId, multiLingualObjectToArray } from "../utils/utils.js";
 import { getMultiLingual } from "../utils/context.js";
-import { currentMulmoScriptVersion } from "../utils/const.js";
+import { currentMulmoScriptVersion } from "../types/const.js";
 import { translateApiKeyMissingError, hasCause, agentGenerationError, translateAction, multiLingualFileTarget } from "../utils/error_cause.js";
 import { getOutputMultilingualFilePath, mkdir, writingMessage, hashSHA256 } from "../utils/file.js";
 import { translateSystemPrompt, translatePrompts } from "../utils/prompt.js";

package/lib/agents/image_genai_agent.js

@@ -1,9 +1,9 @@
 import fs from "fs";
 import { GraphAILogger } from "graphai";
-import { provider2ImageAgent } from "../utils/provider2agent.js";
+import { provider2ImageAgent } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentIncorrectAPIKeyError, agentGenerationError, agentInvalidResponseError, imageAction, imageFileTarget, hasCause, getGenAIErrorReason, resultify, } from "../utils/error_cause.js";
 import { getAspectRatio } from "../utils/utils.js";
-import { ASPECT_RATIOS, PRO_ASPECT_RATIOS } from "../utils/const.js";
+import { ASPECT_RATIOS, PRO_ASPECT_RATIOS } from "../types/const.js";
 import { GoogleGenAI, PersonGeneration } from "@google/genai";
 const getGeminiContents = (prompt, referenceImages) => {
     const contents = [{ text: prompt }];

package/lib/agents/image_openai_agent.js

@@ -2,7 +2,7 @@ import fs from "fs";
 import path from "path";
 import { GraphAILogger } from "graphai";
 import OpenAI, { toFile, AuthenticationError, RateLimitError, APIError } from "openai";
-import { provider2ImageAgent, gptImages } from "../utils/provider2agent.js";
+import { provider2ImageAgent, gptImages } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentGenerationError, openAIAgentGenerationError, agentIncorrectAPIKeyError, agentAPIRateLimitError, agentInvalidResponseError, imageAction, imageFileTarget, } from "../utils/error_cause.js";
 // https://platform.openai.com/docs/guides/image-generation
 export const imageOpenaiAgent = async ({ namedInputs, params, config, }) => {

package/lib/agents/image_replicate_agent.js

@@ -3,7 +3,7 @@ import { GraphAILogger } from "graphai";
 import Replicate from "replicate";
 import { getAspectRatio } from "./movie_replicate_agent.js";
 import { apiKeyMissingError, agentIncorrectAPIKeyError, agentGenerationError, agentInvalidResponseError, imageAction, imageFileTarget, hasCause, } from "../utils/error_cause.js";
-import { provider2ImageAgent } from "../utils/provider2agent.js";
+import { provider2ImageAgent } from "../types/provider2agent.js";
 export const imageReplicateAgent = async ({ namedInputs, params, config, }) => {
     const { prompt, referenceImages } = namedInputs;
     const { canvasSize } = params;

package/lib/agents/lipsync_replicate_agent.js

@@ -1,7 +1,7 @@
 import { readFileSync, existsSync } from "fs";
 import { GraphAILogger } from "graphai";
 import Replicate from "replicate";
-import { provider2LipSyncAgent } from "../utils/provider2agent.js";
+import { provider2LipSyncAgent } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentGenerationError, agentFileNotExistError, imageAction, movieFileTarget, audioFileTarget, hasCause, } from "../utils/error_cause.js";
 export const lipSyncReplicateAgent = async ({ namedInputs, params, config, }) => {
     const { movieFile, audioFile, imageFile } = namedInputs;

package/lib/agents/movie_genai_agent.js

@@ -3,8 +3,8 @@ import { GraphAILogger, sleep } from "graphai";
 import { GoogleGenAI, PersonGeneration } from "@google/genai";
 import { apiKeyMissingError, agentGenerationError, agentInvalidResponseError, imageAction, movieFileTarget, videoDurationTarget, hasCause, } from "../utils/error_cause.js";
 import { getAspectRatio } from "../utils/utils.js";
-import { ASPECT_RATIOS } from "../utils/const.js";
-import { getModelDuration, provider2MovieAgent } from "../utils/provider2agent.js";
+import { ASPECT_RATIOS } from "../types/const.js";
+import { getModelDuration, provider2MovieAgent } from "../types/provider2agent.js";
 const pollUntilDone = async (ai, operation) => {
     const response = { operation };
     while (!response.operation.done) {

package/lib/agents/movie_replicate_agent.js

@@ -2,7 +2,7 @@ import { readFileSync } from "fs";
 import { GraphAILogger } from "graphai";
 import Replicate from "replicate";
 import { apiKeyMissingError, agentGenerationError, agentInvalidResponseError, imageAction, movieFileTarget, videoDurationTarget, unsupportedModelTarget, } from "../utils/error_cause.js";
-import { provider2MovieAgent, getModelDuration } from "../utils/provider2agent.js";
+import { provider2MovieAgent, getModelDuration } from "../types/provider2agent.js";
 async function generateMovie(model, apiKey, prompt, imagePath, aspectRatio, duration) {
     const replicate = new Replicate({
         auth: apiKey,

package/lib/agents/sound_effect_replicate_agent.js

@@ -1,7 +1,7 @@
 import { readFileSync } from "fs";
 import { GraphAILogger } from "graphai";
 import Replicate from "replicate";
-import { provider2SoundEffectAgent } from "../utils/provider2agent.js";
+import { provider2SoundEffectAgent } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentGenerationError, imageAction, movieFileTarget, hasCause } from "../utils/error_cause.js";
 export const soundEffectReplicateAgent = async ({ namedInputs, params, config }) => {
     const { prompt, movieFile } = namedInputs;

package/lib/agents/tts_elevenlabs_agent.js

@@ -1,5 +1,5 @@
 import { GraphAILogger } from "graphai";
-import { provider2TTSAgent } from "../utils/provider2agent.js";
+import { provider2TTSAgent } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentVoiceLimitReachedError, agentIncorrectAPIKeyError, agentGenerationError, audioAction, audioFileTarget, } from "../utils/error_cause.js";
 export const ttsElevenlabsAgent = async ({ namedInputs, params, config, }) => {
     const { text } = namedInputs;

package/lib/agents/tts_gemini_agent.js

@@ -1,6 +1,6 @@
 import { GraphAILogger } from "graphai";
 import { GoogleGenAI } from "@google/genai";
-import { provider2TTSAgent } from "../utils/provider2agent.js";
+import { provider2TTSAgent } from "../types/provider2agent.js";
 import { agentIncorrectAPIKeyError, apiKeyMissingError, agentGenerationError, audioAction, audioFileTarget, getGenAIErrorReason, } from "../utils/error_cause.js";
 import { pcmToMp3 } from "../utils/ffmpeg_utils.js";
 const getPrompt = (text, instructions) => {

package/lib/agents/tts_kotodama_agent.js

@@ -1,5 +1,5 @@
 import { GraphAILogger } from "graphai";
-import { provider2TTSAgent } from "../utils/provider2agent.js";
+import { provider2TTSAgent } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentIncorrectAPIKeyError, agentGenerationError, audioAction, audioFileTarget } from "../utils/error_cause.js";
 export const ttsKotodamaAgent = async ({ namedInputs, params, config, }) => {
     const { text } = namedInputs;

package/lib/agents/tts_openai_agent.js

@@ -1,10 +1,10 @@
 import { GraphAILogger } from "graphai";
 import OpenAI, { AuthenticationError, RateLimitError } from "openai";
-import { provider2TTSAgent } from "../utils/provider2agent.js";
+import { provider2TTSAgent } from "../types/provider2agent.js";
 import { apiKeyMissingError, agentIncorrectAPIKeyError, agentAPIRateLimitError, agentGenerationError, audioAction, audioFileTarget, } from "../utils/error_cause.js";
 export const ttsOpenaiAgent = async ({ namedInputs, params, config, }) => {
     const { text } = namedInputs;
-    const { model, voice, suppressError, instructions } = params;
+    const { model, voice, suppressError, instructions, speed } = params;
     const { apiKey, baseURL } = config ?? {};
     if (!apiKey) {
         throw new Error("OpenAI API key is required (OPENAI_API_KEY)", {
@@ -21,6 +21,9 @@ export const ttsOpenaiAgent = async ({ namedInputs, params, config, }) => {
         if (instructions) {
             tts_options["instructions"] = instructions;
         }
+        if (speed) {
+            tts_options["speed"] = speed;
+        }
         GraphAILogger.log("ttsOptions", tts_options);
         const response = await openai.audio.speech.create(tts_options);
         const buffer = Buffer.from(await response.arrayBuffer());

package/lib/cli/commands/bundle/handler.js

@@ -1,6 +1,6 @@
 import { mulmoViewerBundle, audio, images, translate } from "../../../actions/index.js";
 import { initializeContext } from "../../helpers.js";
-import { bundleTargetLang } from "../../../utils/const.js";
+import { bundleTargetLang } from "../../../types/const.js";
 export const handler = async (argv) => {
     const context = await initializeContext(argv);
     if (!context) {

package/lib/cli/commands/movie/builder.js

@@ -1,5 +1,5 @@
 import { commonOptions } from "../../common.js";
-import { languages } from "../../../utils/const.js";
+import { languages } from "../../../types/const.js";
 export const builder = (yargs) => commonOptions(yargs)
     .option("a", {
     alias: "audiodir",

package/lib/cli/commands/pdf/builder.js

@@ -1,5 +1,5 @@
 import { commonOptions } from "../../common.js";
-import { pdf_modes, pdf_sizes } from "../../../utils/const.js";
+import { pdf_modes, pdf_sizes } from "../../../types/const.js";
 export const builder = (yargs) => commonOptions(yargs)
     .option("i", {
     alias: "imagedir",

package/lib/cli/commands/tool/scripting/builder.js

@@ -1,4 +1,4 @@
-import { llm } from "../../../../utils/provider2agent.js";
+import { llm } from "../../../../types/provider2agent.js";
 import { getAvailablePromptTemplates } from "../../../../utils/file.js";
 const availableTemplateNames = getAvailablePromptTemplates().map((template) => template.filename);
 export const builder = (yargs) => {

package/lib/cli/commands/tool/scripting/handler.d.ts

@@ -1,5 +1,5 @@
 import { ToolCliArgs } from "../../../../types/cli_types.js";
-import type { LLM } from "../../../../utils/provider2agent.js";
+import type { LLM } from "../../../../types/provider2agent.js";
 export declare const handler: (argv: ToolCliArgs<{
     o?: string;
     b?: string;

package/lib/cli/commands/tool/scripting/handler.js

@@ -1,5 +1,5 @@
 import { getBaseDirPath, getFullPath } from "../../../../utils/file.js";
-import { outDirName, cacheDirName } from "../../../../utils/const.js";
+import { outDirName, cacheDirName } from "../../../../types/const.js";
 import { getUrlsIfNeeded, selectTemplate } from "../../../../utils/inquirer.js";
 import { createMulmoScriptFromUrl, createMulmoScriptFromFile } from "../../../../tools/create_mulmo_script_from_url.js";
 import { createMulmoScriptInteractively } from "../../../../tools/create_mulmo_script_interactively.js";

package/lib/cli/commands/tool/story_to_script/builder.js

@@ -1,6 +1,6 @@
 import { getAvailablePromptTemplates } from "../../../../utils/file.js";
-import { llm } from "../../../../utils/provider2agent.js";
-import { storyToScriptGenerateMode } from "../../../../utils/const.js";
+import { llm } from "../../../../types/provider2agent.js";
+import { storyToScriptGenerateMode } from "../../../../types/const.js";
 const availableTemplateNames = getAvailablePromptTemplates().map((template) => template.filename);
 export const builder = (yargs) => {
     return yargs

package/lib/cli/commands/tool/story_to_script/handler.d.ts

@@ -1,5 +1,5 @@
 import { ToolCliArgs } from "../../../../types/cli_types.js";
-import type { LLM } from "../../../../utils/provider2agent.js";
+import type { LLM } from "../../../../types/provider2agent.js";
 export declare const handler: (argv: ToolCliArgs<{
     o?: string;
     b?: string;

package/lib/cli/commands/tool/story_to_script/handler.js

@@ -3,7 +3,7 @@ import { setGraphAILogger } from "../../../../cli/helpers.js";
 import { storyToScript } from "../../../../tools/story_to_script.js";
 import { mulmoStoryboardSchema } from "../../../../types/schema.js";
 import { getBaseDirPath, getFullPath, readAndParseJson } from "../../../../utils/file.js";
-import { outDirName } from "../../../../utils/const.js";
+import { outDirName } from "../../../../types/const.js";
 export const handler = async (argv) => {
     const { v: verbose, s: filename, file, o: outdir, b: basedir, beats_per_scene, llm, llm_model, mode } = argv;
     let { t: template } = argv;

package/lib/cli/common.js

@@ -1,4 +1,4 @@
-import { languages } from "../utils/const.js";
+import { languages } from "../types/const.js";
 export const commonOptions = (yargs) => {
     return yargs
         .option("o", {

package/lib/cli/helpers.js

@@ -4,7 +4,7 @@ import path from "path";
 import clipboardy from "clipboardy";
 import { getBaseDirPath, getFullPath, getOutputStudioFilePath, resolveDirPath, mkdir, getOutputMultilingualFilePath, generateTimestampedFileName, } from "../utils/file.js";
 import { isHttp } from "../utils/utils.js";
-import { outDirName, imageDirName, audioDirName } from "../utils/const.js";
+import { outDirName, imageDirName, audioDirName } from "../types/const.js";
 import { MulmoStudioContextMethods } from "../methods/mulmo_studio_context.js";
 import { translate } from "../actions/translate.js";
 import { initializeContextFromFiles } from "../utils/context.js";

package/lib/data/scriptTemplates.js

@@ -484,7 +484,7 @@ export const scriptTemplates = [
         ],
         filename: "html",
         htmlImageParams: {
-            model: "claude-3-7-sonnet-20250219",
+            model: "claude-sonnet-4-5-20250929",
             provider: "anthropic",
         },
         lang: "en",
@@ -987,7 +987,7 @@ export const scriptTemplates = [
         ],
         filename: "presentation",
         htmlImageParams: {
-            model: "claude-3-7-sonnet-20250219",
+            model: "claude-sonnet-4-5-20250929",
             provider: "anthropic",
         },
         lang: "en",

package/lib/data/templateDataSet.js

@@ -49,7 +49,7 @@ export const templateDataSet = {
         "```",
     html: "Another LLM will generate actual slides from the prompt and data for each beat. Adding optional data would help it to generate more compelling slide. Mention the reference in one of beats, if it exists. The valid type of reference is 'article', 'paper', 'image', 'video', 'audio'. Use the JSON below as a template.\n" +
         "```JSON\n" +
-        '{"$mulmocast":{"version":"1.1","credit":"closing"},"references":[{"url":"https://www.somegreatwebsite.com/article/123","title":"Title of the article we are referencing","type":"[TYPE OF ARTICLE: article, paper, image, video, audio]"}],"title":"[TITLE: Brief, engaging title for the topic]","htmlImageParams":{"provider":"anthropic","model":"claude-3-7-sonnet-20250219"},"lang":"en","beats":[{"text":"[NARRATION: Narration for the beat.]","htmlPrompt":{"prompt":"[PROMPT to create appropriate HTML page for the beat.]"}},{"text":"[NARRATION: Narration for the beat.]","htmlPrompt":{"prompt":"[PROMPT to create appropriate HTML page for the beat with the data.]","data":{"description":"DATA TO BE PRESENTED IN THIS BEAT (in any format)]","net_income":{"Q2 FY2024":320,"Q3 FY2024":333,"Q4 FY2024":350},"unit":"USD (Million)"}}}],"canvasSize":{"width":1536,"height":1024}}\n' +
+        '{"$mulmocast":{"version":"1.1","credit":"closing"},"references":[{"url":"https://www.somegreatwebsite.com/article/123","title":"Title of the article we are referencing","type":"[TYPE OF ARTICLE: article, paper, image, video, audio]"}],"title":"[TITLE: Brief, engaging title for the topic]","htmlImageParams":{"provider":"anthropic","model":"claude-sonnet-4-5-20250929"},"lang":"en","beats":[{"text":"[NARRATION: Narration for the beat.]","htmlPrompt":{"prompt":"[PROMPT to create appropriate HTML page for the beat.]"}},{"text":"[NARRATION: Narration for the beat.]","htmlPrompt":{"prompt":"[PROMPT to create appropriate HTML page for the beat with the data.]","data":{"description":"DATA TO BE PRESENTED IN THIS BEAT (in any format)]","net_income":{"Q2 FY2024":320,"Q3 FY2024":333,"Q4 FY2024":350},"unit":"USD (Million)"}}}],"canvasSize":{"width":1536,"height":1024}}\n' +
         "```",
     image_prompt: "Another AI will generate images for each beat based on the image prompt of that beat. Movie prompts must be written in English. Mention the reference in one of beats, if it exists. Use the JSON below as a template.\n" +
         "```JSON\n" +

package/lib/index.common.d.ts

@@ -1,6 +1,6 @@
 export * from "./types/index.js";
-export * from "./utils/provider2agent.js";
-export * from "./utils/const.js";
+export * from "./types/provider2agent.js";
+export * from "./types/const.js";
 export * from "./utils/string.js";
 export * from "./utils/utils.js";
 export * from "./utils/prompt.js";

package/lib/index.common.js

@@ -1,7 +1,7 @@
 // Entry point for universal code
 export * from "./types/index.js";
-export * from "./utils/provider2agent.js";
-export * from "./utils/const.js";
+export * from "./types/provider2agent.js";
+export * from "./types/const.js";
 export * from "./utils/string.js";
 export * from "./utils/utils.js";
 export * from "./utils/prompt.js";

package/lib/mcp/server.js

@@ -9,7 +9,7 @@ import { fileURLToPath } from "url";
 import { GraphAILogger } from "graphai";
 import { audio, images, movie, captions, pdf } from "../actions/index.js";
 import { initializeContext, runTranslateIfNeeded } from "../cli/helpers.js";
-import { outDirName } from "../utils/const.js";
+import { outDirName } from "../types/const.js";
 import { resolveDirPath, mkdir, generateTimestampedFileName } from "../utils/file.js";
 import { MulmoScriptMethods } from "../methods/index.js";
 dotenv.config({ quiet: true });

package/lib/methods/mulmo_presentation_style.d.ts

@@ -177,19 +177,19 @@ export declare const MulmoPresentationStyleMethods: {
     };
     getSoundEffectAgentInfo(presentationStyle: MulmoPresentationStyle, beat: MulmoBeat): {
         agentName: string;
-        defaultModel: import("../utils/provider2agent.js").ReplicateModel;
+        defaultModel: import("../types/provider2agent.js").ReplicateModel;
         keyName: string;
-        models: import("../utils/provider2agent.js").ReplicateModel[];
-        modelParams: Record<import("../utils/provider2agent.js").ReplicateModel, {
+        models: import("../types/provider2agent.js").ReplicateModel[];
+        modelParams: Record<import("../types/provider2agent.js").ReplicateModel, {
             identifier?: `${string}/${string}:${string}`;
         }>;
     };
     getLipSyncAgentInfo(presentationStyle: MulmoPresentationStyle, beat: MulmoBeat): {
         agentName: string;
-        defaultModel: import("../utils/provider2agent.js").ReplicateModel;
+        defaultModel: import("../types/provider2agent.js").ReplicateModel;
         keyName: string;
-        models: import("../utils/provider2agent.js").ReplicateModel[];
-        modelParams: Record<import("../utils/provider2agent.js").ReplicateModel, {
+        models: import("../types/provider2agent.js").ReplicateModel[];
+        modelParams: Record<import("../types/provider2agent.js").ReplicateModel, {
             identifier?: `${string}/${string}:${string}` | `${string}/${string}`;
             video?: string;
             audio: string;

package/lib/methods/mulmo_presentation_style.js

@@ -6,7 +6,7 @@
 import { isNull } from "graphai";
 import { userAssert } from "../utils/utils.js";
 import { text2ImageProviderSchema, text2HtmlImageProviderSchema, text2MovieProviderSchema, text2SpeechProviderSchema, mulmoCanvasDimensionSchema, mulmoTransitionSchema, } from "../types/schema.js";
-import { provider2ImageAgent, provider2MovieAgent, provider2LLMAgent, provider2SoundEffectAgent, provider2LipSyncAgent, defaultProviders, } from "../utils/provider2agent.js";
+import { provider2ImageAgent, provider2MovieAgent, provider2LLMAgent, provider2SoundEffectAgent, provider2LipSyncAgent, defaultProviders, } from "../types/provider2agent.js";
 const defaultTextSlideStyles = [
     '*,*::before,*::after{box-sizing:border-box}body,h1,h2,h3,h4,p,figure,blockquote,dl,dd{margin:0}ul[role="list"],ol[role="list"]{list-style:none}html:focus-within{scroll-behavior:smooth}body{min-height:100vh;text-rendering:optimizeSpeed;line-height:1.5}a:not([class]){text-decoration-skip-ink:auto}img,picture{max-width:100%;display:block}input,button,textarea,select{font:inherit}@media(prefers-reduced-motion:reduce){html:focus-within{scroll-behavior:auto}*,*::before,*::after{animation-duration:.01ms !important;animation-iteration-count:1 !important;transition-duration:.01ms !important;scroll-behavior:auto !important}}',
     "body { margin: 60px; margin-top: 40px; color:#333; font-size: 30px; font-family: Arial, sans-serif; box-sizing: border-box; height: 100vh }",

package/lib/methods/mulmo_studio_context.d.ts

@@ -4,7 +4,7 @@
  * Works in both Node.js and modern browsers.
  */
 import { BeatSessionType, MulmoStudioContext, SessionProgressCallback, SessionType, MulmoBeat, SpeechOptions } from "../types/index.js";
-import { provider2TTSAgent } from "../utils/provider2agent.js";
+import { provider2TTSAgent } from "../types/provider2agent.js";
 export declare const addSessionProgressCallback: (cb: SessionProgressCallback) => void;
 export declare const removeSessionProgressCallback: (cb: SessionProgressCallback) => void;
 export declare const MulmoStudioContextMethods: {

package/lib/tools/story_to_script.d.ts

@@ -1,5 +1,5 @@
 import { MulmoStoryboard, StoryToScriptGenerateMode } from "../types/index.js";
-import type { LLM } from "../utils/provider2agent.js";
+import type { LLM } from "../types/provider2agent.js";
 export declare const storyToScript: ({ story, beatsPerScene, templateName, outdir, fileName, llm, llmModel, generateMode, }: {
     story: MulmoStoryboard;
     beatsPerScene: number;

package/lib/tools/story_to_script.js

@@ -10,7 +10,7 @@ import { graphDataScriptGeneratePrompt, sceneToBeatsPrompt, storyToScriptInfoPro
 import { fileWriteAgent } from "@graphai/vanilla_node_agents";
 import validateSchemaAgent from "../agents/validate_schema_agent.js";
 import { llmPair } from "../utils/utils.js";
-import { storyToScriptGenerateMode } from "../utils/const.js";
+import { storyToScriptGenerateMode } from "../types/const.js";
 import { cliLoadingPlugin } from "../utils/plugins.js";
 const vanillaAgents = agents.default ?? agents;
 const createValidatedScriptGraphData = ({ systemPrompt, prompt, schema, llmAgent, llmModel, maxTokens, }) => {

package/lib/types/agent.d.ts

@@ -111,6 +111,7 @@ export type TTSAgentParams = {
 export type OpenAITTSAgentParams = TTSAgentParams & {
     instructions: string;
     model: string;
+    speed: number;
 };
 export type NijivoiceTTSAgentParams = TTSAgentParams & {
     speed: number;

package/lib/types/const.d.ts

@@ -0,0 +1,15 @@
+export declare const currentMulmoScriptVersion = "1.1";
+export declare const outDirName = "output";
+export declare const audioDirName = "audio";
+export declare const imageDirName = "images";
+export declare const cacheDirName = "cache";
+export declare const pdf_modes: string[];
+export declare const pdf_sizes: string[];
+export declare const languages: string[];
+export declare const storyToScriptGenerateMode: {
+    stepWise: string;
+    oneStep: string;
+};
+export declare const bundleTargetLang: string[];
+export declare const ASPECT_RATIOS: string[];
+export declare const PRO_ASPECT_RATIOS: string[];

package/lib/types/const.js

@@ -0,0 +1,15 @@
+export const currentMulmoScriptVersion = "1.1";
+export const outDirName = "output";
+export const audioDirName = "audio";
+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", "fr", "es", "de", "zh-CN", "zh-TW", "ko", "it", "pt", "ar", "hi"];
+export const storyToScriptGenerateMode = {
+    stepWise: "step_wise",
+    oneStep: "one_step",
+};
+export const bundleTargetLang = ["ja", "en"];
+export const ASPECT_RATIOS = ["1:1", "9:16", "16:9"];
+export const PRO_ASPECT_RATIOS = ["1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9"];

package/lib/types/provider2agent.d.ts

@@ -0,0 +1,191 @@
+export declare const provider2TTSAgent: {
+    nijivoice: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        keyName: string;
+    };
+    openai: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultModel: string;
+        defaultVoice: string;
+        keyName: string;
+        baseURLKeyName: string;
+    };
+    google: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        keyName: string;
+    };
+    gemini: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultModel: string;
+        defaultVoice: string;
+        models: string[];
+        keyName: string;
+    };
+    elevenlabs: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultModel: string;
+        models: string[];
+        keyName: string;
+    };
+    kotodama: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultVoice: string;
+        defaultDecoration: string;
+        keyName: string;
+    };
+    mock: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultModel: string;
+        models: string[];
+    };
+};
+export declare const gptImages: string[];
+export declare const provider2ImageAgent: {
+    openai: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+        keyName: string;
+        baseURLKeyName: string;
+    };
+    google: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+        keyName: string;
+    };
+    replicate: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+        keyName: string;
+    };
+    mock: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+        keyName: string;
+    };
+};
+export type ReplicateModel = `${string}/${string}`;
+export declare const provider2MovieAgent: {
+    replicate: {
+        agentName: string;
+        defaultModel: ReplicateModel;
+        keyName: string;
+        models: string[];
+        modelParams: Record<ReplicateModel, {
+            durations: number[];
+            start_image: string | undefined;
+            last_image?: string;
+            price_per_sec: number;
+        }>;
+    };
+    google: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+        keyName: string;
+        modelParams: {
+            "veo-3.1-generate-preview": {
+                durations: number[];
+            };
+            "veo-3.0-generate-001": {
+                durations: number[];
+            };
+            "veo-2.0-generate-001": {
+                durations: number[];
+            };
+        };
+    };
+    mock: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+        keyName: string;
+        modelParams: {};
+    };
+};
+export declare const provider2SoundEffectAgent: {
+    replicate: {
+        agentName: string;
+        defaultModel: ReplicateModel;
+        keyName: string;
+        models: ReplicateModel[];
+        modelParams: Record<ReplicateModel, {
+            identifier?: `${string}/${string}:${string}`;
+        }>;
+    };
+};
+export declare const provider2LipSyncAgent: {
+    replicate: {
+        agentName: string;
+        defaultModel: ReplicateModel;
+        keyName: string;
+        models: ReplicateModel[];
+        modelParams: Record<ReplicateModel, {
+            identifier?: `${string}/${string}:${string}` | `${string}/${string}`;
+            video?: string;
+            audio: string;
+            image?: string;
+        }>;
+    };
+};
+export declare const provider2LLMAgent: {
+    readonly openai: {
+        readonly agentName: "openAIAgent";
+        readonly defaultModel: "gpt-5";
+        readonly keyName: "OPENAI_API_KEY";
+        readonly baseURLKeyName: "OPENAI_BASE_URL";
+        readonly max_tokens: 8192;
+        readonly models: readonly ["gpt-5", "gpt-5-nano", "gpt-5-mini", "gpt-4.1", "gpt-4.1-mini", "gpt-4.1-nano", "o3", "o3-mini", "o3-pro", "o1", "o1-pro", "gpt-4o", "gpt-4o-mini"];
+    };
+    readonly anthropic: {
+        readonly agentName: "anthropicAgent";
+        readonly defaultModel: "claude-sonnet-4-5-20250929";
+        readonly max_tokens: 8192;
+        readonly models: readonly ["claude-opus-4-1-20250805", "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-sonnet-4-5-20250929", "claude-haiku-4-5-20251001"];
+        readonly keyName: "ANTHROPIC_API_KEY";
+        readonly apiKeyNameOverride: "ANTHROPIC_API_TOKEN";
+    };
+    readonly gemini: {
+        readonly agentName: "geminiAgent";
+        readonly defaultModel: "gemini-2.5-flash";
+        readonly max_tokens: 8192;
+        readonly models: readonly ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite", "gemini-2.0-flash"];
+        readonly keyName: "GEMINI_API_KEY";
+    };
+    readonly groq: {
+        readonly agentName: "groqAgent";
+        readonly defaultModel: "llama-3.1-8b-instant";
+        readonly keyName: "GROQ_API_KEY";
+        readonly max_tokens: 4096;
+        readonly models: readonly ["llama-3.1-8b-instant", "llama-3.3-70b-versatile", "deepseek-r1-distill-llama-70b", "openai/gpt-oss-120b", "openai/gpt-oss-20b"];
+    };
+    readonly mock: {
+        readonly agentName: "mediaMockAgent";
+        readonly defaultModel: "mock";
+        readonly max_tokens: 4096;
+        readonly models: readonly ["mock"];
+    };
+};
+export declare const defaultProviders: {
+    tts: keyof typeof provider2TTSAgent;
+    text2image: keyof typeof provider2ImageAgent;
+    text2movie: keyof typeof provider2MovieAgent;
+    text2Html: keyof typeof provider2LLMAgent;
+    llm: keyof typeof provider2LLMAgent;
+    soundEffect: keyof typeof provider2SoundEffectAgent;
+    lipSync: keyof typeof provider2LipSyncAgent;
+};
+export declare const llm: (keyof typeof provider2LLMAgent)[];
+export type LLM = keyof typeof provider2LLMAgent;
+export declare const htmlLLMProvider: string[];
+export declare const getModelDuration: (provider: keyof typeof provider2MovieAgent, model: string, movieDuration?: number) => number | undefined;

package/lib/types/provider2agent.js

@@ -0,0 +1,326 @@
+// node & browser
+export const provider2TTSAgent = {
+    nijivoice: {
+        agentName: "ttsNijivoiceAgent",
+        hasLimitedConcurrency: true,
+        keyName: "NIJIVOICE_API_KEY",
+    },
+    openai: {
+        agentName: "ttsOpenaiAgent",
+        hasLimitedConcurrency: false,
+        defaultModel: "gpt-4o-mini-tts",
+        defaultVoice: "shimmer",
+        keyName: "OPENAI_API_KEY",
+        baseURLKeyName: "OPENAI_BASE_URL",
+    },
+    google: {
+        agentName: "ttsGoogleAgent",
+        hasLimitedConcurrency: false,
+        keyName: "GEMINI_API_KEY",
+    },
+    gemini: {
+        agentName: "ttsGeminiAgent",
+        hasLimitedConcurrency: false,
+        defaultModel: "gemini-2.5-flash-preview-tts",
+        defaultVoice: "Kore",
+        models: ["gemini-2.5-flash-preview-tts", "gemini-2.5-pro-preview-tts"],
+        keyName: "GEMINI_API_KEY",
+    },
+    elevenlabs: {
+        agentName: "ttsElevenlabsAgent",
+        hasLimitedConcurrency: true,
+        defaultModel: "eleven_multilingual_v2",
+        // Models | ElevenLabs Documentation
+        // https://elevenlabs.io/docs/models
+        models: ["eleven_multilingual_v2", "eleven_turbo_v2_5", "eleven_turbo_v2", "eleven_flash_v2_5", "eleven_flash_v2"],
+        keyName: "ELEVENLABS_API_KEY",
+    },
+    kotodama: {
+        agentName: "ttsKotodamaAgent",
+        hasLimitedConcurrency: true,
+        defaultVoice: "Atla",
+        defaultDecoration: "neutral",
+        keyName: "KOTODAMA_API_KEY",
+    },
+    mock: {
+        agentName: "mediaMockAgent",
+        hasLimitedConcurrency: true,
+        defaultModel: "mock-model",
+        models: ["mock-model"],
+    },
+};
+export const gptImages = ["gpt-image-1.5", "gpt-image-1", "gpt-image-1-mini"];
+export const provider2ImageAgent = {
+    openai: {
+        agentName: "imageOpenaiAgent",
+        defaultModel: "gpt-image-1",
+        models: ["dall-e-3", ...gptImages],
+        keyName: "OPENAI_API_KEY",
+        baseURLKeyName: "OPENAI_BASE_URL",
+    },
+    google: {
+        agentName: "imageGenAIAgent",
+        defaultModel: "gemini-2.5-flash-image",
+        models: ["imagen-4.0-generate-preview-06-06", "imagen-4.0-ultra-generate-preview-06-06", "gemini-2.5-flash-image", "gemini-3-pro-image-preview"],
+        keyName: "GEMINI_API_KEY",
+    },
+    replicate: {
+        agentName: "imageReplicateAgent",
+        defaultModel: "bytedance/seedream-4",
+        models: ["bytedance/seedream-4", "qwen/qwen-image"],
+        keyName: "REPLICATE_API_TOKEN",
+    },
+    mock: {
+        agentName: "mediaMockAgent",
+        defaultModel: "mock-model",
+        models: ["mock-model"],
+        keyName: "",
+    },
+};
+export const provider2MovieAgent = {
+    replicate: {
+        agentName: "movieReplicateAgent",
+        defaultModel: "bytedance/seedance-1-lite",
+        keyName: "REPLICATE_API_TOKEN",
+        models: [
+            "bytedance/seedance-1-lite",
+            "bytedance/seedance-1-pro",
+            "kwaivgi/kling-v1.6-pro",
+            "kwaivgi/kling-v2.1",
+            "kwaivgi/kling-v2.1-master",
+            "google/veo-2",
+            "google/veo-3",
+            "google/veo-3-fast",
+            "minimax/video-01",
+            "minimax/hailuo-02",
+            "minimax/hailuo-02-fast",
+            "pixverse/pixverse-v4.5",
+            "wan-video/wan-2.2-i2v-fast",
+            "wan-video/wan-2.2-t2v-fast",
+        ],
+        modelParams: {
+            "bytedance/seedance-1-lite": {
+                durations: [5, 10],
+                start_image: "image",
+                last_image: "last_frame_image",
+                price_per_sec: 0.036, // in USD
+            },
+            "bytedance/seedance-1-pro": {
+                durations: [5, 10],
+                start_image: "image",
+                last_image: "last_frame_image",
+                price_per_sec: 0.15,
+            },
+            "kwaivgi/kling-v1.6-pro": {
+                durations: [5, 10],
+                start_image: "start_image",
+                price_per_sec: 0.095,
+            },
+            "kwaivgi/kling-v2.1": {
+                durations: [5, 10],
+                start_image: "start_image",
+                price_per_sec: 0.05,
+            },
+            "kwaivgi/kling-v2.1-master": {
+                durations: [5, 10],
+                start_image: "start_image",
+                price_per_sec: 0.28,
+            },
+            "google/veo-2": {
+                durations: [5, 6, 7, 8],
+                start_image: "image",
+                price_per_sec: 0.5,
+            },
+            "google/veo-3": {
+                durations: [8],
+                start_image: "image",
+                price_per_sec: 0.75,
+            },
+            "google/veo-3-fast": {
+                durations: [8],
+                start_image: "image",
+                price_per_sec: 0.4,
+            },
+            "minimax/video-01": {
+                durations: [6],
+                start_image: "first_frame_image",
+                price_per_sec: 0.5,
+            },
+            "minimax/hailuo-02": {
+                durations: [6], // NOTE: 10 for only 720p
+                start_image: "first_frame_image",
+                price_per_sec: 0.08,
+            },
+            "minimax/hailuo-02-fast": {
+                durations: [6, 10], // NOTE: 512P
+                start_image: "first_frame_image",
+                price_per_sec: 0.0166,
+            },
+            "pixverse/pixverse-v4.5": {
+                durations: [5, 8],
+                start_image: "image",
+                last_image: "last_frame_image",
+                price_per_sec: 0.12,
+            },
+            "wan-video/wan-2.2-i2v-fast": {
+                durations: [5],
+                start_image: "image",
+                price_per_sec: 0.012,
+            },
+            "wan-video/wan-2.2-t2v-fast": {
+                durations: [5],
+                start_image: undefined,
+                price_per_sec: 0.012,
+            },
+        },
+    },
+    google: {
+        agentName: "movieGenAIAgent",
+        defaultModel: "veo-2.0-generate-001",
+        models: ["veo-2.0-generate-001", "veo-3.0-generate-001", "veo-3.1-generate-preview"],
+        keyName: "GEMINI_API_KEY",
+        modelParams: {
+            "veo-3.1-generate-preview": {
+                durations: [4, 6, 8],
+            },
+            "veo-3.0-generate-001": {
+                durations: [4, 6, 8],
+            },
+            "veo-2.0-generate-001": {
+                durations: [5, 6, 7, 8],
+            },
+        },
+    },
+    mock: {
+        agentName: "mediaMockAgent",
+        defaultModel: "mock-model",
+        models: ["mock-model"],
+        keyName: "",
+        modelParams: {},
+    },
+};
+export const provider2SoundEffectAgent = {
+    replicate: {
+        agentName: "soundEffectReplicateAgent",
+        defaultModel: "zsxkib/mmaudio",
+        keyName: "REPLICATE_API_TOKEN",
+        models: ["zsxkib/mmaudio"],
+        modelParams: {
+            "zsxkib/mmaudio": {
+                identifier: "zsxkib/mmaudio:62871fb59889b2d7c13777f08deb3b36bdff88f7e1d53a50ad7694548a41b484",
+            },
+        },
+    },
+};
+export const provider2LipSyncAgent = {
+    replicate: {
+        agentName: "lipSyncReplicateAgent",
+        defaultModel: "bytedance/omni-human",
+        keyName: "REPLICATE_API_TOKEN",
+        models: ["bytedance/latentsync", "tmappdev/lipsync", "bytedance/omni-human"],
+        modelParams: {
+            "bytedance/latentsync": {
+                identifier: "bytedance/latentsync:637ce1919f807ca20da3a448ddc2743535d2853649574cd52a933120e9b9e293",
+                video: "video",
+                audio: "audio",
+            },
+            "tmappdev/lipsync": {
+                identifier: "tmappdev/lipsync:c54ce2fe673ea59b857b91250b3d71a2cd304a78f2370687632805c8405fbf4c",
+                video: "video_input",
+                audio: "audio_input",
+            },
+            "bytedance/omni-human": {
+                identifier: "bytedance/omni-human",
+                image: "image",
+                audio: "audio",
+                price_per_sec: 0.14,
+            },
+            /* NOTE: This model does not work with large base64 urls.
+            "sync/lipsync-2": {
+              video: "video",
+              audio: "audio",
+            },
+            */
+            /* NOTE: This model does not work well for some unknown reason.
+            "kwaivgi/kling-lip-sync": {
+              video: "video_url",
+              audio: "audio_file",
+            },
+            */
+        },
+    },
+};
+// : Record<LLM, { agent: string; defaultModel: string; max_tokens: number }>
+export const provider2LLMAgent = {
+    openai: {
+        agentName: "openAIAgent",
+        defaultModel: "gpt-5",
+        keyName: "OPENAI_API_KEY",
+        baseURLKeyName: "OPENAI_BASE_URL",
+        max_tokens: 8192,
+        models: [
+            "gpt-5",
+            "gpt-5-nano",
+            "gpt-5-mini",
+            "gpt-4.1",
+            "gpt-4.1-mini",
+            "gpt-4.1-nano",
+            "o3",
+            "o3-mini",
+            "o3-pro",
+            "o1",
+            "o1-pro",
+            "gpt-4o",
+            "gpt-4o-mini",
+        ],
+    },
+    anthropic: {
+        agentName: "anthropicAgent",
+        defaultModel: "claude-sonnet-4-5-20250929",
+        max_tokens: 8192,
+        models: ["claude-opus-4-1-20250805", "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-sonnet-4-5-20250929", "claude-haiku-4-5-20251001"],
+        keyName: "ANTHROPIC_API_KEY",
+        apiKeyNameOverride: "ANTHROPIC_API_TOKEN",
+        // GraphAI is currently using ANTHROPIC_API_KEY, but the official name is ANTHROPIC_API_TOKEN.
+    },
+    gemini: {
+        agentName: "geminiAgent",
+        defaultModel: "gemini-2.5-flash",
+        max_tokens: 8192,
+        models: ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite", "gemini-2.0-flash"],
+        keyName: "GEMINI_API_KEY",
+    },
+    groq: {
+        agentName: "groqAgent",
+        defaultModel: "llama-3.1-8b-instant",
+        keyName: "GROQ_API_KEY",
+        max_tokens: 4096,
+        models: ["llama-3.1-8b-instant", "llama-3.3-70b-versatile", "deepseek-r1-distill-llama-70b", "openai/gpt-oss-120b", "openai/gpt-oss-20b"],
+    },
+    mock: {
+        agentName: "mediaMockAgent",
+        defaultModel: "mock",
+        max_tokens: 4096,
+        models: ["mock"],
+    },
+};
+export const defaultProviders = {
+    tts: "openai",
+    text2image: "openai",
+    text2movie: "replicate",
+    text2Html: "openai",
+    llm: "openai",
+    soundEffect: "replicate",
+    lipSync: "replicate",
+};
+export const llm = Object.keys(provider2LLMAgent);
+export const htmlLLMProvider = ["openai", "anthropic", "mock"];
+export const getModelDuration = (provider, model, movieDuration) => {
+    const modelParams = provider2MovieAgent[provider]?.modelParams;
+    const { durations } = modelParams[model];
+    if (durations && movieDuration) {
+        const largerDurations = durations.filter((d) => d >= movieDuration);
+        return largerDurations.length > 0 ? largerDurations[0] : durations[durations.length - 1];
+    }
+    return durations?.[0];
+};

package/lib/types/schema.js

@@ -1,6 +1,6 @@
 import { z } from "zod";
-import { htmlLLMProvider, provider2TTSAgent, provider2ImageAgent, provider2MovieAgent, defaultProviders, provider2SoundEffectAgent, } from "../utils/provider2agent.js";
-import { currentMulmoScriptVersion } from "../utils/const.js";
+import { htmlLLMProvider, provider2TTSAgent, provider2ImageAgent, provider2MovieAgent, defaultProviders, provider2SoundEffectAgent } from "./provider2agent.js";
+import { currentMulmoScriptVersion } from "./const.js";
 import { mulmoVideoFilterSchema } from "./schema_video_filter.js";
 // Re-export video filter schema
 export { mulmoVideoFilterSchema } from "./schema_video_filter.js";

package/lib/types/type.d.ts

@@ -1,7 +1,7 @@
 import { type CallbackFunction } from "graphai";
 import { langSchema, localizedTextSchema, mulmoBeatSchema, mulmoScriptSchema, mulmoStudioSchema, mulmoStudioBeatSchema, mulmoStoryboardSchema, mulmoStoryboardSceneSchema, mulmoStudioMultiLingualSchema, mulmoStudioMultiLingualArraySchema, mulmoStudioMultiLingualDataSchema, mulmoStudioMultiLingualFileSchema, speakerDictionarySchema, speakerSchema, mulmoSpeechParamsSchema, mulmoImageParamsSchema, mulmoImageParamsImagesValueSchema, mulmoImageParamsImagesSchema, mulmoFillOptionSchema, mulmoTransitionSchema, mulmoVideoFilterSchema, mulmoMovieParamsSchema, mulmoSoundEffectParamsSchema, mulmoLipSyncParamsSchema, textSlideParamsSchema, speechOptionsSchema, speakerDataSchema, mulmoCanvasDimensionSchema, mulmoPromptTemplateSchema, mulmoPromptTemplateFileSchema, text2ImageProviderSchema, text2HtmlImageProviderSchema, text2MovieProviderSchema, text2SpeechProviderSchema, mulmoPresentationStyleSchema, multiLingualTextsSchema, mulmoImageAssetSchema, mulmoMermaidMediaSchema, mulmoTextSlideMediaSchema, mulmoMarkdownMediaSchema, mulmoImageMediaSchema, mulmoChartMediaSchema, mediaSourceSchema, mediaSourceMermaidSchema, mulmoSessionStateSchema, mulmoOpenAIImageModelSchema, mulmoGoogleImageModelSchema, mulmoGoogleMovieModelSchema, mulmoReplicateMovieModelSchema, mulmoImagePromptMediaSchema } from "./schema.js";
-import { pdf_modes, pdf_sizes, storyToScriptGenerateMode } from "../utils/const.js";
-import type { LLM } from "../utils/provider2agent.js";
+import { pdf_modes, pdf_sizes, storyToScriptGenerateMode } from "./const.js";
+import type { LLM } from "./provider2agent.js";
 import { z } from "zod";
 export type LANG = z.infer<typeof langSchema>;
 export type MulmoBeat = z.infer<typeof mulmoBeatSchema>;

package/lib/utils/utils.d.ts

@@ -5,7 +5,7 @@
  */
 import type { ConfigDataDictionary, DefaultConfigData } from "graphai";
 import { MulmoBeat, MulmoStudioBeat, MulmoStudioMultiLingual, MulmoStudioMultiLingualData } from "../types/index.js";
-import { type LLM } from "./provider2agent.js";
+import { type LLM } from "../types/provider2agent.js";
 export declare const llmPair: (_llm?: LLM, _model?: string) => {
     agent: "mediaMockAgent" | "openAIAgent" | "anthropicAgent" | "geminiAgent" | "groqAgent";
     model: string;

package/lib/utils/utils.js

@@ -3,7 +3,7 @@
  * (No Node.js built-ins like fs, path, dotenv, etc.)
  * Works in both Node.js and modern browsers.
  */
-import { provider2LLMAgent, provider2TTSAgent, provider2ImageAgent, provider2MovieAgent, provider2SoundEffectAgent, provider2LipSyncAgent, } from "./provider2agent.js";
+import { provider2LLMAgent, provider2TTSAgent, provider2ImageAgent, provider2MovieAgent, provider2SoundEffectAgent, provider2LipSyncAgent, } from "../types/provider2agent.js";
 export const llmPair = (_llm, _model) => {
     const llmKey = _llm ?? "openai";
     const agent = provider2LLMAgent[llmKey]?.agentName ?? provider2LLMAgent.openai.agentName;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "2.1.15",
+  "version": "2.1.16",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -77,7 +77,7 @@
     "@google/genai": "^1.34.0",
     "@graphai/anthropic_agent": "^2.0.12",
     "@graphai/browserless_agent": "^2.0.1",
-    "@graphai/gemini_agent": "^2.0.1",
+    "@graphai/gemini_agent": "^2.0.2",
     "@graphai/groq_agent": "^2.0.2",
     "@graphai/input_agents": "^1.0.2",
     "@graphai/openai_agent": "^2.0.8",
@@ -94,7 +94,7 @@
     "dotenv": "^17.2.3",
     "fluent-ffmpeg": "^2.1.3",
     "graphai": "^2.0.16",
-    "jsdom": "^27.3.0",
+    "jsdom": "^27.4.0",
     "marked": "^17.0.1",
     "mulmocast-vision": "^1.0.8",
     "ora": "^9.0.0",
@@ -102,7 +102,7 @@
     "replicate": "^1.4.0",
     "yaml": "^2.8.2",
     "yargs": "^18.0.0",
-    "zod": "^4.2.1"
+    "zod": "^4.3.5"
   },
   "devDependencies": {
     "@receptron/test_utils": "^2.0.3",
@@ -117,7 +117,7 @@
     "prettier": "^3.7.4",
     "tsx": "^4.21.0",
     "typescript": "^5.9.3",
-    "typescript-eslint": "^8.50.1"
+    "typescript-eslint": "^8.52.0"
   },
   "engines": {
     "node": ">=20.0.0"

package/scripts/templates/html.json

@@ -13,7 +13,7 @@
   "title": "[TITLE: Brief, engaging title for the topic]",
   "htmlImageParams": {
     "provider": "anthropic",
-    "model": "claude-3-7-sonnet-20250219"
+    "model": "claude-sonnet-4-5-20250929"
   },
   "lang": "en",
   "beats": [

package/scripts/templates/presentation.json

@@ -5,7 +5,7 @@
   },
   "htmlImageParams": {
     "provider": "anthropic",
-    "model": "claude-3-7-sonnet-20250219"
+    "model": "claude-sonnet-4-5-20250929"
   },
   "title": "Sample Title",
   "references": [