mulmocast 2.9.0 → 2.9.2

package/lib/utils/provider2agent.d.ts

@@ -0,0 +1,141 @@
+export declare const provider2TTSAgent: {
+    nijivoice: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+    };
+    openai: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultModel: string;
+        defaultVoice: string;
+    };
+    google: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+    };
+    elevenlabs: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultModel: string;
+        models: string[];
+    };
+    mock: {
+        agentName: string;
+        hasLimitedConcurrency: boolean;
+        defaultModel: string;
+        models: string[];
+    };
+};
+export declare const provider2ImageAgent: {
+    openai: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+    };
+    google: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+    };
+    replicate: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+    };
+    mock: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+    };
+};
+export type ReplicateModel = `${string}/${string}`;
+export declare const provider2MovieAgent: {
+    replicate: {
+        agentName: string;
+        defaultModel: ReplicateModel;
+        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[];
+    };
+    mock: {
+        agentName: string;
+        defaultModel: string;
+        models: string[];
+    };
+};
+export declare const provider2SoundEffectAgent: {
+    replicate: {
+        agentName: string;
+        defaultModel: ReplicateModel;
+        models: ReplicateModel[];
+        modelParams: Record<ReplicateModel, {
+            identifier?: `${string}/${string}:${string}`;
+        }>;
+    };
+};
+export declare const provider2LipSyncAgent: {
+    replicate: {
+        agentName: string;
+        defaultModel: ReplicateModel;
+        models: ReplicateModel[];
+        modelParams: Record<ReplicateModel, {
+            identifier?: `${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 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-3-7-sonnet-20250219";
+        readonly max_tokens: 8192;
+        readonly models: readonly ["claude-opus-4-1-20250805", "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-3-7-sonnet-20250219", "claude-3-haiku-20240307"];
+    };
+    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 groq: {
+        readonly agentName: "groqAgent";
+        readonly defaultModel: "llama-3.1-8b-instant";
+        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[];

package/lib/utils/provider2agent.js

@@ -0,0 +1,267 @@
+// node & browser
+export const provider2TTSAgent = {
+    nijivoice: {
+        agentName: "ttsNijivoiceAgent",
+        hasLimitedConcurrency: true,
+    },
+    openai: {
+        agentName: "ttsOpenaiAgent",
+        hasLimitedConcurrency: false,
+        defaultModel: "gpt-4o-mini-tts",
+        defaultVoice: "shimmer",
+    },
+    google: {
+        agentName: "ttsGoogleAgent",
+        hasLimitedConcurrency: false,
+    },
+    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"],
+    },
+    mock: {
+        agentName: "mediaMockAgent",
+        hasLimitedConcurrency: true,
+        defaultModel: "mock-model",
+        models: ["mock-model"],
+    },
+};
+export const provider2ImageAgent = {
+    openai: {
+        agentName: "imageOpenaiAgent",
+        defaultModel: "gpt-image-1",
+        models: ["dall-e-3", "gpt-image-1"],
+    },
+    google: {
+        agentName: "imageGenAIAgent",
+        defaultModel: "gemini-2.5-flash-image-preview",
+        models: ["imagen-3.0-generate-002", "imagen-4.0-generate-preview-06-06", "imagen-4.0-ultra-generate-preview-06-06", "gemini-2.5-flash-image-preview"],
+    },
+    replicate: {
+        agentName: "imageReplicateAgent",
+        defaultModel: "bytedance/seedream-4",
+        models: ["bytedance/seedream-4", "qwen/qwen-image"],
+    },
+    mock: {
+        agentName: "mediaMockAgent",
+        defaultModel: "mock-model",
+        models: ["mock-model"],
+    },
+};
+export const provider2MovieAgent = {
+    replicate: {
+        agentName: "movieReplicateAgent",
+        defaultModel: "bytedance/seedance-1-lite",
+        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-480p-fast",
+            "wan-video/wan-2.2-t2v-480p-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-480p-fast": {
+                durations: [5],
+                start_image: "image",
+                price_per_sec: 0.012,
+            },
+            "wan-video/wan-2.2-t2v-480p-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-preview"],
+    },
+    mock: {
+        agentName: "mediaMockAgent",
+        defaultModel: "mock-model",
+        models: ["mock-model"],
+    },
+};
+export const provider2SoundEffectAgent = {
+    replicate: {
+        agentName: "soundEffectReplicateAgent",
+        defaultModel: "zsxkib/mmaudio",
+        models: ["zsxkib/mmaudio"],
+        modelParams: {
+            "zsxkib/mmaudio": {
+                identifier: "zsxkib/mmaudio:62871fb59889b2d7c13777f08deb3b36bdff88f7e1d53a50ad7694548a41b484",
+            },
+        },
+    },
+};
+export const provider2LipSyncAgent = {
+    replicate: {
+        agentName: "lipSyncReplicateAgent",
+        defaultModel: "bytedance/omni-human",
+        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",
+        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-3-7-sonnet-20250219",
+        max_tokens: 8192,
+        models: ["claude-opus-4-1-20250805", "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-3-7-sonnet-20250219", "claude-3-haiku-20240307"],
+    },
+    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"],
+    },
+    groq: {
+        agentName: "groqAgent",
+        defaultModel: "llama-3.1-8b-instant",
+        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"];

package/lib/utils/text_hash.d.ts

@@ -0,0 +1 @@
+export declare const text2hash: (input: string) => string;

package/lib/utils/text_hash.js

@@ -0,0 +1,4 @@
+import * as crypto from "crypto";
+export const text2hash = (input) => {
+    return crypto.createHash("sha256").update(input).digest("hex");
+};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "2.9.0",
+  "version": "2.9.2",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -25,7 +25,7 @@
   },
   "resolutions": {
     "minimatch": "^10.2.5",
-    "tar": "7.5.20",
+    "tar": "7.5.22",
     "yauzl": "^3.4.0"
   },
   "bin": {
@@ -89,7 +89,7 @@
   "homepage": "https://github.com/receptron/mulmocast-cli#readme",
   "dependencies": {
     "@google-cloud/text-to-speech": "^6.4.1",
-    "@google/genai": "^2.12.0",
+    "@google/genai": "^2.13.0",
     "@graphai/anthropic_agent": "^2.0.12",
     "@graphai/browserless_agent": "^2.0.3",
     "@graphai/gemini_agent": "^2.0.5",
@@ -107,12 +107,12 @@
     "@mulmocast/deck": "^1.1.0",
     "@tavily/core": "^0.7.6",
     "archiver": "^8.0.0",
-    "clipboardy": "^5.3.1",
+    "clipboardy": "^5.3.2",
     "dotenv": "^17.4.2",
     "graphai": "^2.0.18",
     "js-tiktoken": "^1.0.21",
     "jsdom": "^29.1.1",
-    "marked": "^18.0.6",
+    "marked": "^18.0.7",
     "mulmocast-vision": "^1.0.12",
     "ora": "^9.4.1",
     "puppeteer": "^25.3.0",
@@ -128,16 +128,16 @@
     "@types/jsdom": "^28.0.3",
     "@types/yargs": "^17.0.35",
     "cross-env": "^10.1.0",
-    "eslint": "^10.7.0",
+    "eslint": "^10.8.0",
     "eslint-config-prettier": "^10.1.8",
     "eslint-plugin-import": "^2.32.0",
     "eslint-plugin-prettier": "^5.5.6",
     "eslint-plugin-sonarjs": "^4.2.0",
     "globals": "^17.7.0",
-    "prettier": "^3.9.5",
+    "prettier": "^3.9.6",
     "tsx": "^4.23.1",
     "typescript": "6.0.3",
-    "typescript-eslint": "^8.64.0"
+    "typescript-eslint": "^8.65.0"
   },
   "engines": {
     "node": ">=22.0.0"

package/scripts/test/test_voice_over_transition_no_audio.json

@@ -0,0 +1,71 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "lang": "en",
+  "title": "VoiceOver Transition Test",
+  "movieParams": {
+    "transition": {
+      "type": "fade",
+      "duration": 1.0
+    }
+  },
+  "beats": [
+    {
+      "speaker": "Presenter",
+      "text": "",
+      "duration": 3.0,
+      "image": {
+        "type": "image",
+        "source": {
+          "kind": "path",
+          "path": "../../assets/images/mulmocast_credit.png"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "",
+      "duration": 2.0,
+      "image": {
+        "type": "voice_over"
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "",
+      "duration": 3.0,
+      "image": {
+        "type": "textSlide",
+        "slide": {
+          "title": "Fade after a voice_over"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "",
+      "duration": 2.0,
+      "image": {
+        "type": "voice_over"
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "",
+      "duration": 3.0,
+      "movieParams": {
+        "transition": {
+          "type": "wipeleft",
+          "duration": 1.0
+        }
+      },
+      "image": {
+        "type": "textSlide",
+        "slide": {
+          "title": "Wipe after a voice_over"
+        }
+      }
+    }
+  ]
+}

package/lib/utils/sdk_timeout.d.ts

@@ -1,3 +0,0 @@
-export declare const GENAI_REQUEST_TIMEOUT_MS = 120000;
-export declare const REPLICATE_RUN_TIMEOUT_MS = 600000;
-export declare const VIDEO_POLL_TIMEOUT_MS = 1200000;

package/lib/utils/sdk_timeout.js

@@ -1,7 +0,0 @@
-// Timeout budgets for provider SDK calls (OpenAI has its own in openai_client.ts).
-// Generous safety nets against a stalled connection or a never-completing
-// operation — NOT tight deadlines. A stall becomes a rejection so the caller's
-// GraphAI `retry` can recover instead of hanging forever.
-export const GENAI_REQUEST_TIMEOUT_MS = 120_000; // per @google/genai API call (kick-off / each poll)
-export const REPLICATE_RUN_TIMEOUT_MS = 600_000; // whole replicate.run() job (video models run minutes)
-export const VIDEO_POLL_TIMEOUT_MS = 1_200_000; // wall-clock cap for a long-running Veo video operation