@effectnode/media 0.3.0 → 0.5.0

package/dist/backend/movie-backend/agent/agent-backend.js

@@ -1,10 +1,10 @@
 import { existsSync, readFileSync, writeFileSync, renameSync, rmSync, } from "node:fs";
 import { join, dirname } from "node:path";
-import { spawn } from "../process";
+import { spawn } from "../process.js";
 import OpenAI from "openai";
-import { getAgentServerPort } from "../render-media";
-import { workspaceDir, resolveWorkspacePath, walkFiles, classifyFile, ensureDir, movieStudioDataDir, movieStudioStateFile, } from "./workspace";
-import { TOOLS, toolDefinitions, runTool } from "./tools";
+import { getAgentServerPort } from "../render-media.js";
+import { workspaceDir, resolveWorkspacePath, walkFiles, classifyFile, ensureDir, movieStudioDataDir, movieStudioStateFile, } from "./workspace.js";
+import { TOOLS, toolDefinitions, runTool } from "./tools/index.js";
 // Node.js cannot `import` a bare `.txt` file (Bun allowed `with { type: "txt" }`).
 // Read the bundled prompt as UTF-8 text at module load instead.
 const story = readFileSync(new URL("./prompt/story-writer.txt", import.meta.url), "utf-8");

package/dist/backend/movie-backend/agent/prompt/story-writer.txt

@@ -0,0 +1,137 @@
+# Vid Script Builder
+
+## Role & Purpose
+
+You are a short-form video planner and AI Prompt Specialist specializing in humorous, viral content. Your mission is to take user-provided video topics, comedy concepts, or scenarios and convert them directly into structured CSV data designed for Text-to-Image (T2I) and Image-to-Video (I2V) production workflows. Each row represents a single moment of a camera shot, with its own duration (3–20 seconds).
+
+## Workflow & Output Rules
+
+When a user provides a theme, scenario, or prompt request, adhere strictly to the following specifications.
+
+### 0. Defaults — Idea & Art Style (never ask; apply defaults)
+
+- **Do not ask the user clarifying questions.** When a value is not supplied, apply the default below and proceed immediately. Never block on a question.
+
+**Idea (topic / scenario / theme):**
+- Take the idea directly from the user's message. Do not ask the user to elaborate or confirm before writing.
+
+**Art Style:**
+- **Every `t2i` prompt must include an explicit art style.**
+- If the user has NOT supplied an art style, use the default: **photo realistic render**.
+- Use that style consistently across every `t2i` prompt in the whole sequence. If the user supplies a different style, use theirs instead.
+
+### 1. Table Schema (CSV Format)
+
+Every output must include the following four columns:
+
+| Column     | Description                               |
+| ---------- | ----------------------------------------- |
+| `id`       | Sequential moment number (1, 2, 3...)     |
+| `duration` | Duration of this moment in seconds (3–20) |
+| `t2i`      | Text-to-Image Prompt (English)            |
+| `i2v`      | Image-to-Video Prompt (English)           |
+
+### 2. Character Table Schema (CSV Format)
+
+Before the story CSV, first output a separate **character CSV** — one row per recurring character in the story. This is the machine-readable character bible that locks every character's face identity so it stays consistent across all shots.
+
+| Column     | Description                                                                 |
+| ---------- | --------------------------------------------------------------------------- |
+| `id`       | Sequential character number (1, 2, 3...)                                    |
+| `name`     | Character name, matching the name used in the story's t2i/i2v prompts       |
+| `face_t2i` | A dedicated Text-to-Image prompt (English) that renders the character's face as a clean reference portrait |
+
+**`face_t2i` rules:**
+- **Purpose:** one clean, front-facing head-and-shoulders portrait that becomes the character's reference face, so every later shot keeps the same identity.
+- **Face only:** lead with the character name, then the full facial description — age, face shape, skin tone, eye color, hair color/style, eyebrows, nose, lips, and any distinctive feature (scar, glasses, freckles, earring). Do not describe body, outfit, or setting.
+- **Neutral reference:** relaxed, neutral expression, direct-to-camera gaze, mouth closed, head centered, even soft lighting, no dramatic shadow or action.
+- **Style:** reuse the same art style verbatim as the story (the user-supplied style, or the default "photo realistic render" when none was given).
+- **Composition:** head-and-shoulders, centered, symmetrical, sharp focus on the face, plain neutral background.
+- **One row per character.** The `name` must match exactly everywhere it appears in the story CSV.
+
+### 3. Scene Breakdown & Segmentation Constraint
+
+- **Each item is a single moment of a camera shot.** Treat every row as one discrete moment/beat within a shot — not a whole scene.
+- **Break large scenes into many small moments.** When a user provides a broad scene, storyline, or concept, decompose it into as many single-moment rows as possible — never compress a large scene into one row.
+- Each row carries a `duration` between **3 and 20 seconds**, but **prefer small moments**: break the action into the smallest coherent beats and favor short durations (3–6 seconds); use a longer duration only when one continuous camera move or action genuinely requires it.
+- Each moment captures exactly one action beat, one camera setup, and one comedic/punchline step.
+- Maintain narrative continuity across consecutive moments (consistent characters, setting, and progression) so the sequence edits together into one coherent video.
+- Pacing, narrative action, character dialogue, and voiceover in each moment must be calibrated to fit strictly within its stated duration.
+
+### 4. t2i (Text-to-Image) Guidelines — Z-Image-Turbo
+
+- **Language:** English.
+- **Model:** Optimized for **Z-Image-Turbo**. Write prompts as **natural-language sentences**, never comma-separated keyword tags (`(masterpiece), 8k, trending on artstation` fails).
+- **Subject first:** Lead with the subject and its action — Z-Image-Turbo weights earlier tokens more heavily, so never bury the main subject late in the description.
+- **Character consistency (critical across shots):** before generating any prompts, write one fixed, detailed character description (a "character bible") for every recurring character — name, age, face shape, hair color/style, build, outfit, and any distinctive feature (scar, glasses, earring). **Every t2i prompt must repeat the full character bible — the visual description of every character in the cast — verbatim, not only the character performing the action.** Lead with the in-focus subject's description, then append the remaining characters' descriptions unchanged, varying only each character's shot-specific action, expression, and pose. Never reword or omit any character's core appearance details between shots, or character identity will drift across scenes.
+- **One focused paragraph** built in this layered order:
+  1. **Subject & Action** — who/what, doing what, with specific physical details.
+  2. **Environment & Context** — location, time of day, weather, background.
+  3. **Lighting & Atmosphere** — the strongest lever after style.
+  4. **Visual Style** — state the **art style** verbatim (the user-supplied style, or the default "photo realistic render" when none was given) plus camera/lens/film details. Never substitute a different style.
+  5. **Composition** — framing, focus, close-up vs. wide shot.
+- **No negative prompts:** Z-Image-Turbo has no classifier-free guidance, so negative prompts are unsupported. Phrase constraints positively (e.g., "a quiet, deserted street" instead of "no people").
+- **Be specific, avoid vagueness:** replace "a dog" with "a golden retriever puppy with one ear flopped sideways, sitting on a porch step."
+- **Keep quality modifiers minimal:** 2–3 technical tags max (e.g., "85mm f/1.8, shallow depth of field").
+- **On-image text (if any):** wrap literal text in straight quotes, keep it to a short phrase, and specify font style and placement.
+- **Scope:** each prompt must describe a single moment of a camera shot, matching its stated duration.
+
+### 5. i2v (Image-to-Video) Guidelines — LTX-2.3
+
+- **Language:** English.
+- **Core rule — describe the motion, not the image.** The t2i image is the anchor that already fixes the subject, setting, and lighting. Do not re-describe static details; write only what changes: motion, camera movement, and performance.
+- **Motion-first:** lead with action verbs and camera verbs in the first 20–40 words. Describe "how the pixels move," not the subject itself.
+- **Scale motion to the moment's duration:** one main action beat plus one simple camera move for short moments (3–6s); add a second beat or a longer continuous move only when the duration exceeds ~10s. Do not stack simultaneous motions, and avoid complex camera terms (e.g., "rack focus", "dolly zoom") that the distilled model drifts on.
+- **Avoid:** generic phrasing ("make it look nice"), contradictory directions ("fast dramatic zoom but calm and slow"), still-photo phrasing ("preserve identity", "gentle parallax"), and prompts over ~150 words.
+- **Never** include the clip duration or the model name in the prompt text.
+- **Negative prompt:** keep negatives in the dedicated negative field (e.g., `no stiff motion, no flat lighting, no artifacts, no morphing, no flickering`), never inside the positive prompt.
+
+Every i2v prompt must still explicitly include these four components, written as a single cinematic paragraph:
+
+| #   | Component           | Description                                                                               |
+| --- | ------------------- | ----------------------------------------------------------------------------------------- |
+| 1   | Camera Movement     | One specific motion directive (e.g., slow dolly-in, steady hold, snap zoom).              |
+| 2   | Character Name      | Named subject performing the action — name once, then focus on their motion/performance.  |
+| 3   | Character Dialogue  | Spoken lines in quotation marks, treated as an audio/lipsync cue.                         |
+| 4   | Voiceover/Narration | Formatted as `Voiceover: "..."` to capture the punchline, inner monologue, or core humor. |
+
+## Output Format Standard
+
+Output two CSVs: the character CSV first, then the story CSV. Put each inside its own clean CSV code block, each followed immediately by a rendered Markdown preview table.
+
+**Always write both CSVs in full after any update.** Whenever the story changes for any reason (an initial request, an edit to a moment, a new art style, an added or removed scene or character, a re-write, or any user correction), regenerate and output the **complete** character CSV (from `id = 1`) and the **complete** story CSV (from `id = 1` through the final moment) — never output only the changed rows, a partial diff, or a single row. The full CSVs are the single source of truth and must always reflect the entire current story and its full cast.
+
+```csv
+id,name,face_t2i
+1,"Alex","[English T2I Prompt: front-facing facial reference portrait]"
+```
+
+```csv
+id,duration,t2i,i2v
+1,5,"[English T2I Prompt]","[English I2V Prompt: Includes camera movement, character name, dialogue, and voiceover narration structured for this moment's duration]"
+```
+
+### Sample Character Bible (reused across every shot)
+
+> Alex — a 20-year-old male college student with short black hair, round wire-frame glasses, and a faded blue hoodie, always with a nervous, wide-eyed look.
+> Sam — a 40-year-old male professor with receding gray hair, a gray tweed blazer, and thick reading glasses perched on his nose, always with a stern, unamused expression.
+>
+> Repeat both characters' exact descriptions verbatim in every t2i prompt, leading with whoever is in focus and appending the other. Change only each character's action, expression, and pose per moment.
+
+### Sample Character CSV
+
+```csv
+id,name,face_t2i
+1,"Alex","Alex's face — a 20-year-old male with a rounded face shape, fair skin, dark brown eyes, short black hair, round wire-frame glasses, a slightly upturned nose, and thin lips, neutral relaxed expression, direct-to-camera gaze, mouth closed, head centered. Head-and-shoulders portrait, symmetrical, sharp focus on the face, even soft frontal lighting, plain neutral background. Photo realistic render, shot on an 85mm lens at f/5.6, shallow depth of field."
+2,"Sam","Sam's face — a 40-year-old male with an angular face shape, light skin, gray eyes, receding gray hair, thick reading glasses perched on his nose, a straight nose, and thin pressed lips, neutral relaxed expression, direct-to-camera gaze, mouth closed, head centered. Head-and-shoulders portrait, symmetrical, sharp focus on the face, even soft frontal lighting, plain neutral background. Photo realistic render, shot on an 85mm lens at f/5.6, shallow depth of field."
+```
+
+### Sample t2i Prompt Structure (single moment)
+
+> A close-up of Alex, a 20-year-old male college student with short black hair, round wire-frame glasses, and a faded blue hoodie, now with wide, panicked eyes and sweat beading on his forehead, frantically gripping a pen in an exam hall. Behind him, Sam, a 40-year-old male professor with receding gray hair, a gray tweed blazer, and thick reading glasses perched on his nose, watches with a stern, unamused expression. Harsh fluorescent overhead lighting, rows of wooden desks receding into the background, other students blurred in soft focus. Photo realistic render, shot on a 50mm lens at f/2.0, shallow depth of field, tense anxious atmosphere.
+
+### Sample i2v Prompt Structure (single moment)
+
+> Slow dolly-in toward Alex's face as his eyes widen and a bead of sweat rolls down his temple. He grips the pen tighter and whispers urgently: "I'm so dead, I literally haven't opened the textbook!" Voiceover: "Every class has that one top student whose catchphrase is 'I failed,' right before scoring 100%."
+
+> **Negative prompt:** `no stiff motion, no flat lighting, no artifacts, no morphing, no flickering`

package/dist/backend/movie-backend/agent/tools/edit-image.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/edit-image.js

@@ -1,6 +1,6 @@
 import { existsSync, mkdirSync, readFileSync, writeFileSync, realpathSync, } from "node:fs";
 import { join, sep } from "node:path";
-import { isValidProjectId, resolveWorkspacePath, classifyFile, AGENT_UPLOAD_DIR, workspaceDir, } from "../workspace";
+import { isValidProjectId, resolveWorkspacePath, classifyFile, AGENT_UPLOAD_DIR, workspaceDir, } from "../workspace.js";
 import sharp from "sharp";
 /** Default output resolution (px) when the tool is called without one. */
 const DEFAULT_RESOLUTION = 512;

package/dist/backend/movie-backend/agent/tools/get-time.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/grep-files.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/grep-files.js

@@ -1,6 +1,6 @@
 import { existsSync, readFileSync } from "node:fs";
 import { join } from "node:path";
-import { resolveWorkspacePath, workspaceDir, walkFiles } from "../workspace";
+import { resolveWorkspacePath, workspaceDir, walkFiles } from "../workspace.js";
 const tool = {
     name: "grep_files",
     description: "Search for a regex pattern within text files in the workspace.",

package/dist/backend/movie-backend/agent/tools/image-to-video-generation.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/image-to-video-generation.js

@@ -1,6 +1,6 @@
 import { existsSync, mkdirSync, readFileSync, writeFileSync, realpathSync, } from "node:fs";
 import { join } from "node:path";
-import { isValidProjectId, resolveWorkspacePath, classifyFile, AGENT_UPLOAD_DIR, workspaceDir, } from "../workspace";
+import { isValidProjectId, resolveWorkspacePath, classifyFile, AGENT_UPLOAD_DIR, workspaceDir, } from "../workspace.js";
 async function readSSE(response, onEvent) {
     const reader = response.body?.getReader();
     if (!reader)

package/dist/backend/movie-backend/agent/tools/index.d.ts

@@ -1,5 +1,5 @@
-import type { AgentTool, ToolRunContext } from "./types";
-export type { AgentTool, ToolRunContext } from "./types";
+import type { AgentTool, ToolRunContext } from "./types.js";
+export type { AgentTool, ToolRunContext } from "./types.js";
 export declare const TOOLS: AgentTool[];
 /** Build the OpenAI `tools` array from the tool objects. */
 export declare function toolDefinitions(tools: AgentTool[]): any[];

package/dist/backend/movie-backend/agent/tools/index.js

@@ -1,16 +1,16 @@
-import getTimeTool from "./get-time";
-import listFilesTool from "./list-files";
-import readFileTool from "./read-file";
-import writeFileTool from "./write-file";
-import updateFileTool from "./update-file";
-import removeFileTool from "./remove-file";
-import renameFileTool from "./rename-file";
-import grepFilesTool from "./grep-files";
-import showImageTool from "./show-image";
-import editImageTool from "./edit-image";
-import imageToVideoGenerationTool from "./image-to-video-generation";
-import textToVideoGenerationTool from "./text-to-video-generation";
-import stitchVideosTool from "./stitch-videos";
+import getTimeTool from "./get-time.js";
+import listFilesTool from "./list-files.js";
+import readFileTool from "./read-file.js";
+import writeFileTool from "./write-file.js";
+import updateFileTool from "./update-file.js";
+import removeFileTool from "./remove-file.js";
+import renameFileTool from "./rename-file.js";
+import grepFilesTool from "./grep-files.js";
+import showImageTool from "./show-image.js";
+import editImageTool from "./edit-image.js";
+import imageToVideoGenerationTool from "./image-to-video-generation.js";
+import textToVideoGenerationTool from "./text-to-video-generation.js";
+import stitchVideosTool from "./stitch-videos.js";
 export const TOOLS = [
     getTimeTool,
     listFilesTool,

package/dist/backend/movie-backend/agent/tools/list-files.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/list-files.js

@@ -1,5 +1,5 @@
 import { existsSync } from "node:fs";
-import { workspaceDir, walkFiles } from "../workspace";
+import { workspaceDir, walkFiles } from "../workspace.js";
 const tool = {
     name: "list_files",
     description: "List all files in the agent's workspace (recursive).",

package/dist/backend/movie-backend/agent/tools/read-file.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/read-file.js

@@ -1,5 +1,5 @@
 import { existsSync, readFileSync } from "node:fs";
-import { resolveWorkspacePath, classifyFile } from "../workspace";
+import { resolveWorkspacePath, classifyFile } from "../workspace.js";
 const tool = {
     name: "read_file",
     description: "Read the text content of a file in the workspace.",

package/dist/backend/movie-backend/agent/tools/remove-file.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/remove-file.js

@@ -1,5 +1,5 @@
 import { existsSync, rmSync } from "node:fs";
-import { resolveWorkspacePath } from "../workspace";
+import { resolveWorkspacePath } from "../workspace.js";
 const tool = {
     name: "remove_file",
     description: "Delete a file (or directory) in the workspace.",

package/dist/backend/movie-backend/agent/tools/rename-file.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/rename-file.js

@@ -1,6 +1,6 @@
 import { existsSync, renameSync } from "node:fs";
 import { dirname, join } from "node:path";
-import { resolveWorkspacePath } from "../workspace";
+import { resolveWorkspacePath } from "../workspace.js";
 const tool = {
     name: "rename_file",
     description: "Rename a file in the workspace.",

package/dist/backend/movie-backend/agent/tools/show-image.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/show-image.js

@@ -1,5 +1,5 @@
 import { existsSync } from "node:fs";
-import { resolveWorkspacePath, classifyFile } from "../workspace";
+import { resolveWorkspacePath, classifyFile } from "../workspace.js";
 const tool = {
     name: "show_image",
     description: "Display an image from the workspace to the user.",

package/dist/backend/movie-backend/agent/tools/stitch-videos.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/stitch-videos.js

@@ -2,7 +2,7 @@ import { FFmpeg } from "@ffmpeg/ffmpeg";
 import { toBlobURL } from "@ffmpeg/util";
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { join } from "node:path";
-import { isValidProjectId, resolveWorkspacePath, classifyFile, workspaceDir, } from "../workspace";
+import { isValidProjectId, resolveWorkspacePath, classifyFile, workspaceDir, } from "../workspace.js";
 const CORE_BASE = "https://unpkg.com/@ffmpeg/core@0.12.10/dist/esm";
 const tool = {
     name: "stitch_videos",

package/dist/backend/movie-backend/agent/tools/text-to-video-generation.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/text-to-video-generation.js

@@ -1,4 +1,4 @@
-import { isValidProjectId, workspaceDir } from "../workspace";
+import { isValidProjectId, workspaceDir } from "../workspace.js";
 async function readSSE(response, onEvent) {
     const reader = response.body?.getReader();
     if (!reader)

package/dist/backend/movie-backend/agent/tools/update-file.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/update-file.js

@@ -1,6 +1,6 @@
 import { existsSync, readFileSync, writeFileSync } from "node:fs";
 import { dirname } from "node:path";
-import { resolveWorkspacePath, ensureDir } from "../workspace";
+import { resolveWorkspacePath, ensureDir } from "../workspace.js";
 const tool = {
     name: "update_file",
     description: "Append text to the end of a file in the workspace.",

package/dist/backend/movie-backend/agent/tools/write-file.d.ts

@@ -1,3 +1,3 @@
-import type { AgentTool } from "./types";
+import type { AgentTool } from "./types.js";
 declare const tool: AgentTool;
 export default tool;

package/dist/backend/movie-backend/agent/tools/write-file.js

@@ -1,6 +1,6 @@
 import { writeFileSync } from "node:fs";
 import { dirname } from "node:path";
-import { resolveWorkspacePath, ensureDir } from "../workspace";
+import { resolveWorkspacePath, ensureDir } from "../workspace.js";
 const tool = {
     name: "write_file",
     description: "Write (create or overwrite) a file with the given text content.",

package/dist/backend/movie-backend/core.js

@@ -2,14 +2,14 @@ import { appendFileSync, existsSync, mkdirSync, readFileSync, } from "node:fs";
 import { dirname, join } from "node:path";
 import { fileURLToPath } from "node:url";
 import { createServer as createNetServer } from "node:net";
-import { spawn } from "./process";
+import { spawn } from "./process.js";
 const __filename = fileURLToPath(import.meta.url);
 import { homedir } from "node:os";
 import express from "express";
 import cors from "cors";
-import { renderMediaRoutes } from "./render-media";
-import { agentBackend } from "./agent/agent-backend";
-import { generationQueueSetup } from "./generation-queue";
+import { renderMediaRoutes } from "./render-media.js";
+import { agentBackend } from "./agent/agent-backend.js";
+import { generationQueueSetup } from "./generation-queue.js";
 import { createServer } from "node:http";
 // import { readdir } from "node:fs/promises";
 // import { rename } from "node:fs/promises";

package/dist/backend/movie-backend/generation-queue.js

@@ -1,10 +1,10 @@
-import { appendFileSync, existsSync, mkdirSync, readFileSync, writeFileSync, } from "node:fs";
+import { appendFileSync, existsSync, mkdirSync, readFileSync, statSync, writeFileSync, } from "node:fs";
 import { dirname, join } from "node:path";
 import { homedir } from "node:os";
 import { randomUUID } from "node:crypto";
-import { generateAssetImage, generateSceneImage, generateSceneVideo, cancelActiveRender, } from "./render-media";
-import { generateMovieStudioBible } from "./agent/agent-backend";
-import { movieStudioStateFile } from "./agent/workspace";
+import { generateAssetImage, generateSceneImage, generateSceneVideo, cancelActiveRender, } from "./render-media.js";
+import { generateMovieStudioBible } from "./agent/agent-backend.js";
+import { movieStudioStateFile } from "./agent/workspace.js";
 const APP_DATA_DIR = join(homedir(), "media-studio");
 const TASKS_DIR = join(APP_DATA_DIR, "tasks");
 const LOGS_DIR = join(APP_DATA_DIR, "logs");
@@ -38,6 +38,25 @@ function slugify(v) {
 function isValidProjectId(id) {
     return PROJECT_ID_RE.test(id);
 }
+const OUTPUT_DIR = join(APP_DATA_DIR, "output");
+/** Return an existing output file's metadata if already generated, else null. */
+function existingOutput(projectId, filename) {
+    const path = join(OUTPUT_DIR, projectId, filename);
+    if (!existsSync(path))
+        return null;
+    let updatedAt = Date.now();
+    try {
+        updatedAt = statSync(path).mtimeMs;
+    }
+    catch {
+        // Fall back to now — only affects the cache-busting query param.
+    }
+    return {
+        filename,
+        url: `/api/files?path=${encodeURIComponent(path)}`,
+        updatedAt,
+    };
+}
 // In-memory queues, keyed by project id. Persisted to disk on every mutation so
 // the queue survives app restarts and is visible at tasks/:projectId/queue.json.
 const queues = new Map();
@@ -193,17 +212,27 @@ async function runRenderAssets(ctx, characters, places) {
             statusText: `Generating character: ${c?.name || slug}`,
             progress: { current, total },
         });
-        const r = await generateAssetImage(ctx.projectId, "character", slug, prompt, ctx.log);
-        throwIfAborted(ctx.signal);
-        if ("error" in r)
-            throw new Error(r.error);
-        assets.push({
-            kind: "character",
-            slug,
-            filename: r.filename,
-            url: r.url,
-            updatedAt: Date.now(),
-        });
+        // Skip media that's already been generated for this character.
+        const existing = existingOutput(ctx.projectId, `character-${slug}.png`);
+        let entry;
+        if (existing) {
+            ctx.log(`Already generated: character-${slug}.png — skipping\n`);
+            entry = { kind: "character", slug, ...existing };
+        }
+        else {
+            const r = await generateAssetImage(ctx.projectId, "character", slug, prompt, ctx.log);
+            throwIfAborted(ctx.signal);
+            if ("error" in r)
+                throw new Error(r.error);
+            entry = {
+                kind: "character",
+                slug,
+                filename: r.filename,
+                url: r.url,
+                updatedAt: Date.now(),
+            };
+        }
+        assets.push(entry);
         ctx.update({ result: { assets: [...assets] } });
     }
     for (const p of placesList) {
@@ -217,17 +246,26 @@ async function runRenderAssets(ctx, characters, places) {
             statusText: `Generating place: ${p?.name || slug}`,
             progress: { current, total },
         });
-        const r = await generateAssetImage(ctx.projectId, "place", slug, prompt, ctx.log);
-        throwIfAborted(ctx.signal);
-        if ("error" in r)
-            throw new Error(r.error);
-        assets.push({
-            kind: "place",
-            slug,
-            filename: r.filename,
-            url: r.url,
-            updatedAt: Date.now(),
-        });
+        const existing = existingOutput(ctx.projectId, `place-${slug}.png`);
+        let entry;
+        if (existing) {
+            ctx.log(`Already generated: place-${slug}.png — skipping\n`);
+            entry = { kind: "place", slug, ...existing };
+        }
+        else {
+            const r = await generateAssetImage(ctx.projectId, "place", slug, prompt, ctx.log);
+            throwIfAborted(ctx.signal);
+            if ("error" in r)
+                throw new Error(r.error);
+            entry = {
+                kind: "place",
+                slug,
+                filename: r.filename,
+                url: r.url,
+                updatedAt: Date.now(),
+            };
+        }
+        assets.push(entry);
         ctx.update({ result: { assets: [...assets] } });
     }
     ctx.update({ result: { assets } });
@@ -248,16 +286,25 @@ async function runRenderSceneImages(ctx, scenes) {
             statusText: `Generating scene image: ${slug}`,
             progress: { current, total },
         });
-        const r = await generateSceneImage(ctx.projectId, sc, ctx.log);
-        throwIfAborted(ctx.signal);
-        if ("error" in r)
-            throw new Error(r.error);
-        sceneImages.push({
-            slug,
-            filename: r.filename,
-            url: r.url,
-            updatedAt: Date.now(),
-        });
+        const existing = existingOutput(ctx.projectId, `scene-${slug}.png`);
+        let entry;
+        if (existing) {
+            ctx.log(`Already generated: scene-${slug}.png — skipping\n`);
+            entry = { slug, ...existing };
+        }
+        else {
+            const r = await generateSceneImage(ctx.projectId, sc, ctx.log);
+            throwIfAborted(ctx.signal);
+            if ("error" in r)
+                throw new Error(r.error);
+            entry = {
+                slug,
+                filename: r.filename,
+                url: r.url,
+                updatedAt: Date.now(),
+            };
+        }
+        sceneImages.push(entry);
         ctx.update({ result: { sceneImages: [...sceneImages] } });
     }
     ctx.update({ result: { sceneImages } });
@@ -279,16 +326,25 @@ async function runRenderVideos(ctx, uvPath, scenes, characters) {
             statusText: `Generating video: ${slug}`,
             progress: { current, total },
         });
-        const r = await generateSceneVideo(uvPath, ctx.projectId, sc, chars, ctx.log);
-        throwIfAborted(ctx.signal);
-        if ("error" in r)
-            throw new Error(r.error);
-        videos.push({
-            slug,
-            filename: r.filename,
-            url: r.url,
-            updatedAt: Date.now(),
-        });
+        const existing = existingOutput(ctx.projectId, `scene-${slug}.mp4`);
+        let entry;
+        if (existing) {
+            ctx.log(`Already generated: scene-${slug}.mp4 — skipping\n`);
+            entry = { slug, ...existing };
+        }
+        else {
+            const r = await generateSceneVideo(uvPath, ctx.projectId, sc, chars, ctx.log);
+            throwIfAborted(ctx.signal);
+            if ("error" in r)
+                throw new Error(r.error);
+            entry = {
+                slug,
+                filename: r.filename,
+                url: r.url,
+                updatedAt: Date.now(),
+            };
+        }
+        videos.push(entry);
         ctx.update({ result: { videos: [...videos] } });
     }
     ctx.update({ result: { videos } });

package/dist/backend/movie-backend/render-media.js

@@ -2,7 +2,7 @@ import { existsSync, mkdirSync, readFileSync, writeFileSync, realpathSync, readd
 import { randomUUID } from "node:crypto";
 import { homedir } from "node:os";
 import { join, sep } from "node:path";
-import { spawn, whichSync, mimeType } from "./process";
+import { spawn, whichSync, mimeType } from "./process.js";
 // Track the currently active spawn process so it can be cancelled
 let activeProc = null;
 // Track the long-running mlx-vlm server process (separate from render jobs).

package/frontend/index.html

@@ -4,7 +4,14 @@
 <head>
   <meta charset="UTF-8" />
   <meta name="viewport" content="width=device-width, initial-scale=1.0" />
-  <title>effectnode-media</title>
+  <meta name="theme-color" content="#eaf8f6" />
+  <link rel="preconnect" href="https://fonts.googleapis.com" />
+  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
+  <link
+    href="https://fonts.googleapis.com/css2?family=Cormorant:ital,wght@0,300..700;1,300..600&display=swap"
+    rel="stylesheet"
+  />
+  <title>Lambobo Studio</title>
 </head>
 
 <body>