mulmocast 2.1.15 → 2.1.17

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": "2.1.17",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -74,18 +74,18 @@
   "homepage": "https://github.com/receptron/mulmocast-cli#readme",
   "dependencies": {
     "@google-cloud/text-to-speech": "^6.4.0",
-    "@google/genai": "^1.34.0",
+    "@google/genai": "^1.37.0",
     "@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",
     "@graphai/stream_agent_filter": "^2.0.2",
     "@graphai/vanilla": "^2.0.12",
     "@graphai/vanilla_node_agents": "^2.0.4",
-    "@inquirer/input": "^5.0.3",
-    "@inquirer/select": "^5.0.3",
+    "@inquirer/input": "^5.0.4",
+    "@inquirer/select": "^5.0.4",
     "@modelcontextprotocol/sdk": "^1.25.1",
     "@mozilla/readability": "^0.6.0",
     "@tavily/core": "^0.5.11",
@@ -94,15 +94,15 @@
     "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",
-    "puppeteer": "^24.34.0",
+    "puppeteer": "^24.35.0",
     "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",
@@ -112,12 +112,12 @@
     "@types/yargs": "^17.0.35",
     "eslint": "^9.39.2",
     "eslint-config-prettier": "^10.1.8",
-    "eslint-plugin-prettier": "^5.5.4",
+    "eslint-plugin-prettier": "^5.5.5",
     "eslint-plugin-sonarjs": "^3.0.5",
-    "prettier": "^3.7.4",
+    "prettier": "^3.8.0",
     "tsx": "^4.21.0",
     "typescript": "^5.9.3",
-    "typescript-eslint": "^8.50.1"
+    "typescript-eslint": "^8.53.0"
   },
   "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": [

package/scripts/test/test_captions_2.json

@@ -0,0 +1,31 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "lang": "en",
+  "title": "Caption Test",
+  "speechParams": {
+    "speakers": {
+      "Presenter": {
+        "voiceId": "shimmer",
+        "displayName": {
+          "en": "Presenter"
+        }
+      }
+    }
+  },
+  "beats": [
+    {
+      "speaker": "Presenter",
+      "id": "pingpongmov",
+      "text": "This is a local movie with audio.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "../../test/assets/hello.mp4"
+        }
+      }
+    }
+  ]
+}