mulmocast 2.6.5 → 2.6.7

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mulmocast",
-  "version": "2.6.5",
+  "version": "2.6.7",
   "description": "",
   "type": "module",
   "main": "lib/index.node.js",
@@ -24,8 +24,9 @@
     }
   },
   "resolutions": {
-    "minimatch": "^10.2.4",
-    "yauzl": "^3.2.1"
+    "minimatch": "^10.2.5",
+    "tar": "7.5.13",
+    "yauzl": "^3.3.0"
   },
   "bin": {
     "mulmo": "lib/cli/bin.js",
@@ -88,7 +89,7 @@
   "homepage": "https://github.com/receptron/mulmocast-cli#readme",
   "dependencies": {
     "@google-cloud/text-to-speech": "^6.4.0",
-    "@google/genai": "^1.46.0",
+    "@google/genai": "^1.49.0",
     "@graphai/anthropic_agent": "^2.0.12",
     "@graphai/browserless_agent": "^2.0.2",
     "@graphai/gemini_agent": "^2.0.5",
@@ -98,23 +99,23 @@
     "@graphai/stream_agent_filter": "^2.0.3",
     "@graphai/vanilla": "^2.0.12",
     "@graphai/vanilla_node_agents": "^2.0.4",
-    "@inquirer/input": "^5.0.10",
-    "@inquirer/select": "^5.1.2",
-    "@modelcontextprotocol/sdk": "^1.27.1",
+    "@inquirer/input": "^5.0.11",
+    "@inquirer/select": "^5.1.3",
+    "@modelcontextprotocol/sdk": "^1.29.0",
     "@mozilla/readability": "^0.6.0",
     "@tavily/core": "^0.5.11",
     "archiver": "^7.0.1",
     "clipboardy": "^5.3.1",
-    "dotenv": "^17.3.1",
+    "dotenv": "^17.4.2",
     "fluent-ffmpeg": "^2.1.3",
     "graphai": "^2.0.16",
-    "jsdom": "^29.0.1",
-    "marked": "^17.0.5",
+    "jsdom": "^29.0.2",
+    "marked": "^18.0.0",
     "mulmocast-vision": "^1.0.9",
     "ora": "^9.3.0",
     "puppeteer": "^24.40.0",
     "replicate": "^1.4.0",
-    "yaml": "^2.8.2",
+    "yaml": "^2.8.3",
     "yargs": "^18.0.0",
     "zod": "^4.3.6"
   },
@@ -126,15 +127,16 @@
     "@types/jsdom": "^28.0.1",
     "@types/yargs": "^17.0.35",
     "cross-env": "^10.1.0",
-    "eslint": "^10.1.0",
+    "eslint": "^10.2.0",
     "eslint-config-prettier": "^10.1.8",
+    "eslint-plugin-import": "^2.32.0",
     "eslint-plugin-prettier": "^5.5.5",
     "eslint-plugin-sonarjs": "^4.0.2",
-    "globals": "^17.4.0",
-    "prettier": "^3.8.1",
+    "globals": "^17.5.0",
+    "prettier": "^3.8.2",
     "tsx": "^4.21.0",
-    "typescript": "6.0.1-rc",
-    "typescript-eslint": "^8.57.1"
+    "typescript": "6.0.2",
+    "typescript-eslint": "^8.58.1"
   },
   "engines": {
     "node": ">=22.0.0"

package/scripts/test/test_audio_mix.json

@@ -0,0 +1,91 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "title": "Audio Mix Manual Test (fixtures)",
+  "lang": "en",
+  "canvasSize": {
+    "width": 320,
+    "height": 180
+  },
+  "speechParams": {
+    "speakers": {
+      "Presenter": {
+        "provider": "openai",
+        "voiceId": "shimmer"
+      }
+    }
+  },
+  "audioParams": {
+    "bgmVolume": 0,
+    "movieVolume": 0.3
+  },
+  "beats": [
+    {
+      "speaker": "Presenter",
+      "text": "Low tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "Mid tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "High tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    }
+  ]
+}

package/scripts/test/test_audio_mix_beat_vol.json

@@ -0,0 +1,100 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "title": "Audio Mix Beat Volume Test (fixtures)",
+  "lang": "en",
+  "canvasSize": {
+    "width": 320,
+    "height": 180
+  },
+  "speechParams": {
+    "speakers": {
+      "Presenter": {
+        "provider": "openai",
+        "voiceId": "shimmer"
+      }
+    }
+  },
+  "audioParams": {
+    "bgmVolume": 0,
+    "movieVolume": 0.5
+  },
+  "beats": [
+    {
+      "speaker": "Presenter",
+      "text": "Low tone, beat movieVolume 0.1.",
+      "audioParams": {
+        "movieVolume": 0.1
+      },
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "Low tone, script default 0.5.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "Mid tone, beat movieVolume 1.0.",
+      "audioParams": {
+        "movieVolume": 1.0
+      },
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "Mid tone, script default 0.5.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "High tone, beat movieVolume 0.0, should be silent.",
+      "audioParams": {
+        "movieVolume": 0.0
+      },
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "High tone, script default 0.5.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    }
+  ]
+}

package/scripts/test/test_audio_mix_ducking.json

@@ -0,0 +1,91 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "title": "Audio Mix Ducking Test (fixtures)",
+  "lang": "en",
+  "canvasSize": {
+    "width": 320,
+    "height": 180
+  },
+  "speechParams": {
+    "speakers": {
+      "Presenter": {
+        "provider": "openai",
+        "voiceId": "shimmer"
+      }
+    }
+  },
+  "audioParams": {
+    "bgmVolume": 0,
+    "ducking": { "ratio": 0.3 }
+  },
+  "beats": [
+    {
+      "speaker": "Presenter",
+      "text": "Low tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "Mid tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "High tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    }
+  ]
+}

package/scripts/test/test_audio_mix_legacy.json

@@ -0,0 +1,90 @@
+{
+  "$mulmocast": {
+    "version": "1.1"
+  },
+  "title": "Audio Mix Legacy Test (fixtures)",
+  "lang": "en",
+  "canvasSize": {
+    "width": 320,
+    "height": 180
+  },
+  "speechParams": {
+    "speakers": {
+      "Presenter": {
+        "provider": "openai",
+        "voiceId": "shimmer"
+      }
+    }
+  },
+  "audioParams": {
+    "bgmVolume": 0
+  },
+  "beats": [
+    {
+      "speaker": "Presenter",
+      "text": "Low tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_low.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "Mid tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_mid.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "text": "High tone with narration.",
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    },
+    {
+      "speaker": "Presenter",
+      "duration": 3.0,
+      "image": {
+        "type": "movie",
+        "source": {
+          "kind": "path",
+          "path": "fixtures/movie_tone_high.mov"
+        }
+      }
+    }
+  ]
+}

package/scripts/test/test_grok.json

@@ -0,0 +1,57 @@
+{
+  "$mulmocast": { "version": "1.1" },
+  "movieParams": {
+    "provider": "replicate",
+    "model": "xai/grok-imagine-video"
+  },
+  "audioParams": {
+    "bgmVolume": 0
+  },
+  "captionParams": {
+    "lang": "en"
+  },
+  "lang": "en",
+  "beats": [
+    {
+      "text": "Testing xai/grok-imagine-video — text-to-video, image-to-video, and duration varieties.",
+      "image": {
+        "type": "textSlide",
+        "slide": {
+          "title": "xai/grok-imagine-video test"
+        }
+      }
+    },
+    {
+      "id": "grok-t2v-5s",
+      "text": "Text-to-video, 5 seconds: a red fox trots through a snowy forest at dawn, pawprints left behind",
+      "duration": 5,
+      "moviePrompt": "a red fox trots through a snowy forest at dawn, pawprints left behind, soft golden light filtering through pine trees"
+    },
+    {
+      "id": "grok-t2v-8s",
+      "text": "Text-to-video, 8 seconds: timelapse of storm clouds forming over a mountain range",
+      "duration": 8,
+      "moviePrompt": "timelapse of dark storm clouds rolling in over a rugged mountain range, lightning flashes in the distance, dramatic cinematic lighting"
+    },
+    {
+      "id": "grok-t2v-10s",
+      "text": "Text-to-video, 10 seconds: bustling Tokyo crossing at night",
+      "duration": 10,
+      "moviePrompt": "aerial dolly shot over Shibuya crossing at night, neon lights reflecting on wet pavement, crowds of people crossing in all directions"
+    },
+    {
+      "id": "grok-i2v-5s",
+      "text": "Image-to-video, 5 seconds: ocean waves animated from a still image",
+      "duration": 5,
+      "imagePrompt": "a calm tropical beach at sunset, crystal-clear turquoise water, golden sand, photorealistic",
+      "moviePrompt": "gentle waves lapping the shore, palm trees swaying softly in the breeze, camera slowly pushing forward"
+    },
+    {
+      "id": "grok-i2v-8s",
+      "text": "Image-to-video, 8 seconds: city skyline coming to life",
+      "duration": 8,
+      "imagePrompt": "futuristic city skyline at dusk, glass skyscrapers with glowing windows, photorealistic",
+      "moviePrompt": "lights in the skyscraper windows flicker on one by one as night falls, slow upward pan along the tallest tower"
+    }
+  ]
+}

package/scripts/test/test_image_references.json

@@ -0,0 +1,74 @@
+{
+  "$mulmocast": { "version": "1.1" },
+  "imageParams": {
+    "provider": "google",
+    "images": {
+      "character": {
+        "type": "imagePrompt",
+        "prompt": "a young woman with short black hair wearing a red leather jacket and dark sunglasses, standing on a neon-lit Tokyo street at night, photo realistic"
+      },
+      "black_cat": {
+        "type": "imagePrompt",
+        "prompt": "small black cat"
+      }
+    }
+  },
+  "movieParams": {
+    "provider": "google",
+    "model": "veo-3.1-generate-preview"
+  },
+  "lang": "en",
+  "beats": [
+    {
+      "id": "veo31_asset_and_style",
+      "text": "veo-3.1 with both ASSET and STYLE reference images",
+      "moviePrompt": "the woman runs through narrow alleyways, neon signs blurring in the rain",
+      "movieParams": {
+        "model": "veo-3.1-generate-preview",
+        "referenceImages": [
+          { "imageName": "character", "referenceType": "ASSET" },
+          { "imageName": "black_cat", "referenceType": "ASSET" }
+        ]
+      }
+    },
+    {
+      "id": "veo31_replicate_asset_reference",
+      "text": "google/veo-3.1 (replicate) with reference images",
+      "moviePrompt": "the woman walks through a crowded neon-lit Tokyo street at night, rain falling gently around her",
+      "movieParams": {
+        "provider": "replicate",
+        "model": "google/veo-3.1",
+        "referenceImages": [
+          { "imageName": "character", "referenceType": "ASSET" },
+          { "imageName": "black_cat", "referenceType": "ASSET" }
+        ]
+      }
+    },
+    {
+      "id": "kling_v3_omni_asset_reference",
+      "text": "kling-v3-omni-video with character ASSET reference image",
+      "moviePrompt": "the woman walks through a crowded neon-lit Tokyo street at night, rain falling gently around her",
+      "movieParams": {
+        "provider": "replicate",
+        "model": "kwaivgi/kling-v3-omni-video",
+        "referenceImages": [
+          { "imageName": "character", "referenceType": "ASSET" },
+          { "imageName": "black_cat", "referenceType": "ASSET" }
+        ]
+      }
+    },
+    {
+      "id": "grok_r2v_reference",
+      "text": "xai/grok-imagine-r2v with character and cat reference images",
+      "moviePrompt": "the woman walks through a crowded neon-lit Tokyo street at night, rain falling gently around her, the black cat weaving between her feet",
+      "movieParams": {
+        "provider": "replicate",
+        "model": "xai/grok-imagine-r2v",
+        "referenceImages": [
+          { "imageName": "character", "referenceType": "ASSET" },
+          { "imageName": "black_cat", "referenceType": "ASSET" }
+        ]
+      }
+    }
+  ]
+}

package/scripts/test/test_kling_v3.json

@@ -0,0 +1,54 @@
+{
+  "$mulmocast": { "version": "1.1" },
+  "movieParams": {
+    "provider": "replicate",
+    "model": "kwaivgi/kling-v3-video"
+  },
+  "imageParams": {
+    "provider": "google"
+  },
+  "audioParams": {
+    "bgmVolume": 0
+  },
+  "captionParams": {
+    "lang": "en"
+  },
+  "lang": "en",
+  "beats": [
+    {
+      "text": "Testing kwaivgi/kling-v3-video — text-to-video and image-to-video varieties.",
+      "image": {
+        "type": "textSlide",
+        "slide": {
+          "title": "kwaivgi/kling-v3-video test"
+        }
+      }
+    },
+    {
+      "id": "kling-v3-t2v-5s",
+      "text": "Text-to-video, 5 seconds: a red fox trots through a snowy forest at dawn, pawprints left behind",
+      "duration": 5,
+      "moviePrompt": "a red fox trots through a snowy forest at dawn, pawprints left behind, soft golden light filtering through pine trees"
+    },
+    {
+      "id": "kling-v3-t2v-10s",
+      "text": "Text-to-video, 10 seconds: bustling Tokyo crossing at night",
+      "duration": 10,
+      "moviePrompt": "aerial dolly shot over Shibuya crossing at night, neon lights reflecting on wet pavement, crowds of people crossing in all directions"
+    },
+    {
+      "id": "kling-v3-i2v-5s",
+      "text": "Image-to-video, 5 seconds: ocean waves animated from a still image",
+      "duration": 5,
+      "imagePrompt": "a calm tropical beach at sunset, crystal-clear turquoise water, golden sand, photorealistic",
+      "moviePrompt": "gentle waves lapping the shore, palm trees swaying softly in the breeze, camera slowly pushing forward"
+    },
+    {
+      "id": "kling-v3-i2v-10s",
+      "text": "Image-to-video, 10 seconds: city skyline coming to life",
+      "duration": 10,
+      "imagePrompt": "futuristic city skyline at dusk, glass skyscrapers with glowing windows, photorealistic",
+      "moviePrompt": "lights in the skyscraper windows flicker on one by one as night falls, slow upward pan along the tallest tower"
+    }
+  ]
+}