mulmocast 2.17.0 → 2.18.0

package/README.md

@@ -76,6 +76,7 @@ npm install -g mulmocast
 ```
 
 You'll also need to install ffmpeg:
+
 ```bash
 # For macOS with Homebrew
 brew install ffmpeg
@@ -85,11 +86,13 @@ brew install ffmpeg
 ```
 
 You can also use [`Dockerfile`](./Dockerfile) which helps you install the pre-requisits.
+
 ```
 docker build -t mulmo-cli .
 ```
 
 You can use the Docker image like this:
+
 ```
 docker run -e OPENAI_API_KEY=<your_openai_api_key> -it mulmo-cli mulmo tool scripting -i -t children_book -o ./ -s story
 ```
@@ -99,16 +102,19 @@ docker run -e OPENAI_API_KEY=<your_openai_api_key> -it mulmo-cli mulmo tool scri
 Create a `.env` file in your project directory with the following API keys:
 
 ### Required
+
 ```bash
 OPENAI_API_KEY=your_openai_api_key
 ```
 
 #### (Optional) For Google's image and TTS.
+
 ```bash
 GEMINI_API_KEY=your_google_gemini_api_key
 ```
 
 #### (Optional) For AI providers
+
 ```bash
 # For Anthropic Claude (htmlPrompt feature)
 ANTHROPIC_API_TOKEN=your_anthropic_api_token
@@ -117,22 +123,26 @@ ANTHROPIC_API_TOKEN=your_anthropic_api_token
 For htmlPrompt configuration, see [docs/image.md](./docs/image.md#2-htmlprompt).
 
 #### (Optional) For Movie models
+
 ```bash
 REPLICATE_API_TOKEN=your_replicate_api_key
 ```
 
 #### (Optional) For TTS models
+
 ```bash
 # For ElevenLabs TTS
 ELEVENLABS_API_KEY=your_elevenlabs_api_key
 ```
 
 #### (Optional) to access web in mulmo tool
+
 ```bash
 BROWSERLESS_API_TOKEN=your_browserless_api_token # to access web in mulmo tool
 ```
 
 #### (Optional) Local usage tracking
+
 ```bash
 MULMOCAST_DUMP_USAGE=1                 # print JSON usage dump to stdout after each CLI action
 MULMOCAST_DUMP_USAGE=/tmp/usage.json   # ...or write it to a file
@@ -150,6 +160,7 @@ gcloud auth application-default login
 ```
 
 Configure in MulmoScript:
+
 ```json
 {
   "imageParams": {
@@ -161,10 +172,10 @@ Configure in MulmoScript:
 }
 ```
 
-| Parameter | Description | Default |
-|-----------|-------------|---------|
-| `vertexai_project` | Google Cloud Project ID | None (enables Vertex AI mode when set) |
-| `vertexai_location` | Region | `us-central1` |
+| Parameter           | Description             | Default                                |
+| ------------------- | ----------------------- | -------------------------------------- |
+| `vertexai_project`  | Google Cloud Project ID | None (enables Vertex AI mode when set) |
+| `vertexai_location` | Region                  | `us-central1`                          |
 
 For detailed setup instructions, see [Vertex AI Setup Guide](./docs/vertexai_en.md).
 
@@ -188,6 +199,7 @@ LLM_OPENAI_API_VERSION=2025-04-01-preview  # optional
 ```
 
 MulmoScript configuration (same as OpenAI):
+
 ```json
 {
   "imageParams": {
@@ -253,6 +265,7 @@ Priority chain: `config (defaults)` < `template/style` < `script` < `config.over
 > **Note**: `kind: "path"` entries in `mulmo.config.json` are resolved relative to the **script file directory**, not the config file location. This is consistent with all other path resolution in MulmoScript.
 
 Verify the merged result with:
+
 ```bash
 mulmo tool info merged --script <script.json>
 ```
@@ -261,36 +274,44 @@ mulmo tool info merged --script <script.json>
 
 1. Create a MulmoScript JSON file with `mulmo tool scripting`
 2. Generate audio with `mulmo audio`
-3. Generate images with `mulmo images` 
+3. Generate images with `mulmo images`
 4. Create final video with `mulmo movie`
 
 ## Quick Start
+
 ### Step 1: Generate a MulmoScript with interactive mode
+
 - Step 1-1: Run the script generation command
+
   ```bash
   mulmo tool scripting -i -t children_book -o ./ -s story
   ```
+
   This will initiate the script creation process.
-  
+
 - Step 1-2: Interactive conversation with AI
 
   After running the command, you'll engage in an interactive conversation with the AI to create your story script. The AI will guide you through the process and help shape your content based on your inputs. Once completed, a JSON file (like ⁠`story-1746600802426.json`) will be generated.
 
 ### Step 2: Generate both audio and images, then combine into video
+
 ```bash
 mulmo movie {generated_script_file}
 ```
+
 Replace `{generated_script_file}` with the output file from STEP 1, such as `story-1746600802426.json`.
 
-Optionally, you can specify *__clipboard* as the script file name to paste the script from the clipboard, which is generated by other interactive environment such as ChatGPT or Claude.
+Optionally, you can specify _\__clipboard_ as the script file name to paste the script from the clipboard, which is generated by other interactive environment such as ChatGPT or Claude.
 
 ## Quick Start for Ghibli-Style Animation
 
 [![Watch Ghibli-Style Animation Example](https://github.com/receptron/mulmocast-cli/blob/main/docs/images/introducing-mulmocast-new-era-presentations-thumbnail.jpg?raw=true)](https://www.youtube.com/watch?v=gz_Oe5p2n3s)  
-*Click the image above to watch an example of what you can create*
+_Click the image above to watch an example of what you can create_
 
 ### Step 1: Check/Update Environment Setup
+
 Verify your `.env` file contains:
+
 ```bash
 OPENAI_API_KEY=your_openai_api_key
 ```
@@ -298,23 +319,29 @@ OPENAI_API_KEY=your_openai_api_key
 > **Note:** Ensure your OpenAI organization is verified to access the GPT Image models. Visit https://platform.openai.com/settings/organization/general and complete the "Verifications" section.
 
 ### Step 2: Generate a Ghibli-style MulmoScript
+
 ```bash
 mulmo tool scripting -i -t ghibli_comic -o ./ -s story
 ```
+
 This will initiate an interactive conversation with the AI to create your Ghibli-inspired story. Once completed, a JSON file (e.g., `story-1747834931950.json`) will be generated.
 
 ### Step 3: Translate the MulmoScript
+
 ```bash
 mulmo translate {generated_script_file_from_step_2}
 ```
 
 ### Step 4: Generate video with Japanese subtitles
+
 ```bash
 mulmo movie {generated_script_file_from_step_2} -c ja
 ```
+
 The `-c ja` flag adds Japanese subtitles to your video.
 
 When the process completes, the CLI will display the path to your generated video file:
+
 ```bash
 Video created successfully! 14.939 sec
 writing: /Users/username/path/to/output/story-1747834931950__ja.mp4
@@ -344,10 +371,7 @@ You can create a minimal JSON with just beats and complete it with `mulmo tool c
 
 ```json
 {
-  "beats": [
-    { "text": "Hello, welcome to MulmoCast!" },
-    { "text": "This is a simple example." }
-  ]
+  "beats": [{ "text": "Hello, welcome to MulmoCast!" }, { "text": "This is a simple example." }]
 }
 ```
 
@@ -366,7 +390,6 @@ mulmo tool complete my_beats.json -s ani -o my_script.json
 
 This generates a complete MulmoScript with all required fields (canvasSize, speechParams, imageParams, etc.) from the style.
 
-
 ## Generate content from MulmoScript
 
 Mulmo provides several commands to handle different aspects of content creation:
@@ -403,11 +426,11 @@ Payload shape:
 {
   "records": 3,
   "byModel": [
-    { "provider": "openai",     "model": "gpt-4o-mini-tts", "records": 1, "inputChars": 20 },
-    { "provider": "gemini",     "model": "gemini-3.8-flash-tts", "records": 1, "inputTokens": 9, "outputTokens": 59, "totalTokens": 68 },
+    { "provider": "openai", "model": "gpt-4o-mini-tts", "records": 1, "inputChars": 20 },
+    { "provider": "gemini", "model": "gemini-3.8-flash-tts", "records": 1, "inputTokens": 9, "outputTokens": 59, "totalTokens": 68 },
     { "provider": "elevenlabs", "model": "eleven_multilingual_v2", "records": 1, "inputChars": 18 }
   ],
-  "snapshot": [ /* per-call UsageRecord[] */ ]
+  "snapshot": [/* per-call UsageRecord[] */]
 }
 ```
 
@@ -437,7 +460,18 @@ const estimates = estimateUsage(script, { targetLangs: ["ja"] });
 
 Each metric carries a `precision` flag: `"exact"` for values that are deterministic from the script (TTS character counts, tokenized OpenAI prompts, fixed image-output token tables) and `"estimated"` for heuristics (LLM output length, speech duration derived from text). `costUSD` appears when pricing data exists in `modelPricing` (`src/types/provider2agent.ts`), where every price records the date it was last verified. See [docs/api.md § Pre-run estimation](./docs/api.md#pre-run-estimation-estimateusage).
 
+### Strict network mode (scripts you did not write)
+
+HTML-based beats (`html_tailwind`, `markdown`, `textSlide`, `chart`, `mermaid`, `slide`, captions, HTML prompts) and PDF output are rendered in a headless browser that normally has full network access. For a script from an untrusted source, add `--strict-network` (or set `MULMO_STRICT_NETWORK=1`; library callers set `context.strictNetwork`):
+
+```bash
+mulmo images untrusted.json --strict-network
+```
+
+The renderer may then load only files in the script's folder and mulmocast's output folders, `data:` / `blob:` URLs, and the CDNs mulmocast's own templates use (`cdn.tailwindcss.com`, `fonts.googleapis.com`, `fonts.gstatic.com`, and on `cdn.jsdelivr.net` only the mermaid and Chart.js files the templates load). Anything else — other local files, remote images in the script, `fetch`, WebSockets, WebRTC, popups, preconnect hints — is blocked, and each blocked request is logged (`strict network: blocked …`). Files next to the script in the same folder remain readable, so keep an untrusted script in a folder of its own (a script loaded by URL gets the output folders only). Requests to the allowed CDNs still reach those providers, so treat what a page sends there as visible to them. Strict mode guards what the browser loads; files the script names with `kind: "path"` (images, movies, logos, mermaid or remotion code) are read by mulmocast itself and are not restricted yet ([#1622](https://github.com/receptron/mulmocast-cli/issues/1622)). Remotion beats are always guarded this way. The `vision` beat renders in its own browser inside `mulmocast-vision`, which the guard cannot reach, so in strict mode a `vision` beat stops with an error instead of rendering unguarded.
+
 ## Cache and Re-run
+
 When running the same `mulmo` command multiple times, previously generated files are treated as cache. For example, audio or image files will not be regenerated if they already exist.
 
 To force regeneration, delete the old files — including temporary files — under the output directory before re-running the command or use -f (force) option.
@@ -464,7 +498,24 @@ Set a theme once with `slideParams.theme`, then use `"type": "slide"` in each be
 {
   "$mulmocast": { "version": "1.1" },
   "slideParams": {
-    "theme": { "colors": { "bg": "0F172A", "bgCard": "1E293B", "bgCardAlt": "334155", "text": "F8FAFC", "textMuted": "CBD5E1", "textDim": "64748B", "primary": "3B82F6", "accent": "8B5CF6", "success": "22C55E", "warning": "F59E0B", "danger": "EF4444", "info": "14B8A6", "highlight": "EC4899" }, "fonts": { "title": "Georgia", "body": "Calibri", "mono": "Consolas" } }
+    "theme": {
+      "colors": {
+        "bg": "0F172A",
+        "bgCard": "1E293B",
+        "bgCardAlt": "334155",
+        "text": "F8FAFC",
+        "textMuted": "CBD5E1",
+        "textDim": "64748B",
+        "primary": "3B82F6",
+        "accent": "8B5CF6",
+        "success": "22C55E",
+        "warning": "F59E0B",
+        "danger": "EF4444",
+        "info": "14B8A6",
+        "highlight": "EC4899"
+      },
+      "fonts": { "title": "Georgia", "body": "Calibri", "mono": "Consolas" }
+    }
   },
   "beats": [
     {
@@ -492,18 +543,18 @@ For detailed layout specifications and content block reference, see the [Slide D
 
 MulmoCast includes 100 pre-designed CSS styles for markdown slides, organized in 10 categories:
 
-| Category | Description | Example Styles |
-|----------|-------------|----------------|
-| business | Professional corporate designs | corporate-blue, executive-gray, finance-green |
-| tech | Technology and developer themes | cyber-neon, matrix-green, terminal-dark |
-| creative | Artistic and expressive styles | artistic-splash, watercolor-soft, graffiti |
-| minimalist | Clean and simple designs | clean-white, zen-beige, nordic-light |
-| nature | Nature-inspired themes | forest-green, ocean-blue, sunset-orange |
-| dark | Dark mode and elegant themes | charcoal-elegant, midnight-blue, obsidian |
-| colorful | Vibrant and bold colors | vibrant-pink, candy-colors, aurora |
-| vintage | Retro and classic styles | retro-70s, art-deco, typewriter |
-| japanese | Japanese aesthetic designs | washi-paper, sakura-pink, zen-garden |
-| geometric | Pattern-based designs | hexagon-pattern, isometric, bauhaus |
+| Category   | Description                     | Example Styles                                |
+| ---------- | ------------------------------- | --------------------------------------------- |
+| business   | Professional corporate designs  | corporate-blue, executive-gray, finance-green |
+| tech       | Technology and developer themes | cyber-neon, matrix-green, terminal-dark       |
+| creative   | Artistic and expressive styles  | artistic-splash, watercolor-soft, graffiti    |
+| minimalist | Clean and simple designs        | clean-white, zen-beige, nordic-light          |
+| nature     | Nature-inspired themes          | forest-green, ocean-blue, sunset-orange       |
+| dark       | Dark mode and elegant themes    | charcoal-elegant, midnight-blue, obsidian     |
+| colorful   | Vibrant and bold colors         | vibrant-pink, candy-colors, aurora            |
+| vintage    | Retro and classic styles        | retro-70s, art-deco, typewriter               |
+| japanese   | Japanese aesthetic designs      | washi-paper, sakura-pink, zen-garden          |
+| geometric  | Pattern-based designs           | hexagon-pattern, isometric, bauhaus           |
 
 ### Usage
 
@@ -543,12 +594,12 @@ https://github.com/receptron/mulmocast-cli/tree/main/scripts
 
 ## Directory Structure
 
-| Directory         | Description                                                                 |
-|-------------------|-----------------------------------------------------------------------------|
-| `output/`         | Artifacts generated by commands — e.g., `.json`, audio, and video files    |
-| `output/audio/`   | Temporary audio fragments used in the final audio assembly                  |
-| `output/images/`  | Image files generated by the `images` command                               |
-| `output/cache/`   | Internal cache files for various processing steps                           |
+| Directory        | Description                                                             |
+| ---------------- | ----------------------------------------------------------------------- |
+| `output/`        | Artifacts generated by commands — e.g., `.json`, audio, and video files |
+| `output/audio/`  | Temporary audio fragments used in the final audio assembly              |
+| `output/images/` | Image files generated by the `images` command                           |
+| `output/cache/`  | Internal cache files for various processing steps                       |
 
 > These directories are automatically created as needed by the `mulmo` commands.
 
@@ -597,6 +648,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
 ```
 
@@ -621,6 +675,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
   -a, --audiodir           Audio output directory                       [string]
 ```
@@ -646,6 +703,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
   -i, --imagedir           Image output directory                       [string]
 ```
@@ -671,6 +731,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
   -a, --audiodir           Audio output directory                       [string]
   -i, --imagedir           Image output directory                       [string]
@@ -700,6 +763,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
   -i, --imagedir           Image output directory                       [string]
       --pdf_mode           PDF mode
@@ -729,6 +795,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
       --image_width        Image width (e.g., 400px, 50%, auto)         [string]
 ```
@@ -754,6 +823,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
       --image_width        Image width (e.g., 400px, 50%, auto)         [string]
 ```
@@ -779,6 +851,9 @@ Options:
   -g, --grouped            Output all files under output/<basename>/ directory
                                                       [boolean] [default: false]
       --backup             create backup media file   [boolean] [default: false]
+      --strict-network     Let HTML renderers reach only local files and the
+                           CDNs mulmocast's templates use (also
+                           MULMO_STRICT_NETWORK=1)    [boolean] [default: false]
   -p, --presentationStyle  Presentation Style                           [string]
 ```
 
@@ -962,9 +1037,6 @@ Examples:
   mulmo tool info styles --format json
 ```
 
-
-
 ## Contributing
 
 For developers interested in contributing to this project, please see [CONTRIBUTING.md](./CONTRIBUTING.md).
-

package/lib/actions/captions.js

@@ -99,7 +99,7 @@ const generateBeatCaptions = async (beat, context, index) => {
             styles: (mergedCaptionParams.styles ?? []).join(";\n"),
             bottomOffset: `${mergedCaptionParams.bottomOffset ?? 0}`,
         });
-        await renderHTMLToImage(htmlData, imagePath, canvasSize.width, canvasSize.height, false, true);
+        await renderHTMLToImage(htmlData, imagePath, canvasSize.width, canvasSize.height, false, true, MulmoStudioContextMethods.getRenderNetworkOptions(context));
         return {
             file: imagePath,
             startAt: beatStartAt + introPadding + beatDuration * cumulativeRatios[subIndex],

package/lib/actions/image_agents.d.ts

@@ -93,5 +93,6 @@ export declare const htmlImageGeneratorAgent: (namedInputs: {
     file: string;
     canvasSize: MulmoCanvasDimension;
     htmlText: string;
+    context: MulmoStudioContext;
 }) => Promise<void>;
 export {};

package/lib/actions/image_agents.js

@@ -292,6 +292,6 @@ export const imagePluginAgent = async (namedInputs) => {
     }
 };
 export const htmlImageGeneratorAgent = async (namedInputs) => {
-    const { file, canvasSize, htmlText } = namedInputs;
-    await renderHTMLToImage(htmlText, file, canvasSize.width, canvasSize.height);
+    const { file, canvasSize, htmlText, context } = namedInputs;
+    await renderHTMLToImage(htmlText, file, canvasSize.width, canvasSize.height, false, false, MulmoStudioContextMethods.getRenderNetworkOptions(context));
 };

package/lib/actions/images.d.ts

@@ -275,11 +275,13 @@ export declare const beat_graph_data: {
                 file: string;
                 canvasSize: import("../types/type.js").MulmoCanvasDimension;
                 htmlText: string;
+                context: MulmoStudioContext;
             }) => Promise<void>;
             inputs: {
                 htmlText: string;
                 canvasSize: string;
                 file: string;
+                context: string;
             };
         };
         imageGenerator: {

package/lib/actions/images.js

@@ -139,6 +139,7 @@ export const beat_graph_data = {
                 htmlText: ":htmlReader.htmlText",
                 canvasSize: ":context.presentationStyle.canvasSize",
                 file: ":preprocessor.htmlImageFile",
+                context: ":context",
             },
         },
         imageGenerator: {

package/lib/actions/pdf.js

@@ -1,11 +1,15 @@
 import fs from "fs";
 import path from "path";
+import os from "os";
+import crypto from "crypto";
 import puppeteer from "puppeteer";
 import { GraphAILogger, sleep } from "graphai";
 import { MulmoPresentationStyleMethods } from "../methods/index.js";
 import { localizedText, isHttp } from "../utils/utils.js";
 import { getOutputPdfFilePath, writingMessage, getHTMLFile, mulmoCreditPath } from "../utils/file.js";
 import { interpolate } from "../utils/html_render.js";
+import { guardRenderPage } from "../utils/render_network_guard.js";
+import { strictNetworkLaunchArgs, withRenderContentSecurityPolicy } from "../utils/render_network_policy.js";
 import { safeFetch, FETCH_DOWNLOAD_TIMEOUT_MS } from "../utils/fetch.js";
 import { MulmoStudioContextMethods } from "../methods/mulmo_studio_context.js";
 const isCI = process.env.CI === "true";
@@ -138,17 +142,34 @@ export const pdfFilePath = (context, pdfMode) => {
     return getOutputPdfFilePath(fileDirs.outDirPath, studio.filename, pdfMode, lang);
 };
 const PDF_CONTENT_TIMEOUT_MS = 60000;
+// Strict mode navigates to a file because the guard's WebRTC removal never reaches a setContent document.
+const loadPdfPage = async (page, html, { strictNetwork, allowedFileRoots }) => {
+    if (!strictNetwork) {
+        await page.setContent(html, { waitUntil: "domcontentloaded", timeout: PDF_CONTENT_TIMEOUT_MS });
+        return;
+    }
+    const tmpFile = path.join(os.tmpdir(), `mulmocast_pdf_${crypto.randomUUID()}.html`);
+    await guardRenderPage(page, [tmpFile, ...(allowedFileRoots ?? [])]);
+    fs.writeFileSync(tmpFile, withRenderContentSecurityPolicy(html));
+    try {
+        await page.goto(`file://${tmpFile}`, { waitUntil: "domcontentloaded", timeout: PDF_CONTENT_TIMEOUT_MS });
+    }
+    finally {
+        fs.rmSync(tmpFile, { force: true });
+    }
+};
 const generatePDF = async (context, pdfMode, pdfSize) => {
     const outputPdfPath = pdfFilePath(context, pdfMode);
     const html = await generatePDFHTML(context, pdfMode, pdfSize);
     const canvasSize = MulmoPresentationStyleMethods.getCanvasSize(context.presentationStyle);
     const pdfOptions = createPDFOptions(pdfSize, pdfMode, canvasSize);
+    const network = MulmoStudioContextMethods.getRenderNetworkOptions(context);
     const browser = await puppeteer.launch({
-        args: isCI ? ["--no-sandbox"] : [],
+        args: [...(isCI ? ["--no-sandbox"] : []), ...(network.strictNetwork ? strictNetworkLaunchArgs() : [])],
     });
     try {
         const page = await browser.newPage();
-        await page.setContent(html, { waitUntil: "domcontentloaded", timeout: PDF_CONTENT_TIMEOUT_MS });
+        await loadPdfPage(page, html, network);
         await sleep(1000);
         await page.pdf({
             path: outputPdfPath,

package/lib/agents/movie_genai_agent.d.ts

@@ -1,5 +1,29 @@
 import type { AgentFunction, AgentFunctionInfo } from "graphai";
 import type { AgentBufferResult, GenAIImageAgentConfig, GoogleMovieAgentParams, MovieAgentInputs } from "../types/agent.js";
+import { type GeminiOmniImage, type GeminiOmniVideoRequest } from "../utils/gemini_omni_video.js";
+type GeminiOmniInteraction = {
+    id?: string;
+    output_video?: {
+        data?: string;
+        mime_type?: string;
+    };
+};
+export type GeminiOmniVideoClient = {
+    interactions: {
+        create: (params: GeminiOmniVideoRequest, options?: {
+            timeout?: number;
+        }) => Promise<GeminiOmniInteraction>;
+    };
+};
+export type GeminiOmniVideoInput = {
+    model: string;
+    prompt: string;
+    aspectRatio: string;
+    requestedSec: number;
+    firstFrame?: GeminiOmniImage;
+    lastFrame?: GeminiOmniImage;
+};
+export declare const generateGeminiOmniVideo: (ai: GeminiOmniVideoClient, input: GeminiOmniVideoInput) => Promise<Buffer>;
 export declare const movieGenAIAgent: AgentFunction<GoogleMovieAgentParams, AgentBufferResult, MovieAgentInputs, GenAIImageAgentConfig>;
 declare const movieGenAIAgentInfo: AgentFunctionInfo;
 export default movieGenAIAgentInfo;