Merge branch 'master' of https://git.yasue.org/ken/RandomVideoMaker

src/musicspot_generator/images.ts

@@ -5,7 +5,8 @@ import fs from 'fs/promises';
 import { logger } from '../lib/logger';
 import { callOpenAI } from '../lib/openai';
 import { callLMStudio } from '../lib/lmstudio';
-import { generateImage as generateFaceImage } from '../lib/image-generator-face';
+// import { generateImage as generateFaceImage } from '../lib/image-generator-face'; // Removed
+import { generateImage } from '../lib/image-generator'; // Added
 
 dotenv.config();
 
@@ -43,8 +44,8 @@ interface Server {
     name: string;
 }
 
-const DEFAULT_SIZE: Size = { width: 720, height: 1280 };
+const DEFAULT_SIZE: Size = { width: 1280, height: 720 };
-const FOLDER = process.argv[2] || process.env.MUSICSPOT_FOLDER || 'zagreb';
+const FOLDER = process.argv[2] || process.env.MUSICSPOT_FOLDER || 'infinitydance';
 const FOLDER_SAFE = FOLDER.replace(/[/\\?%*:|"<>]/g, '_');
 const FACE_SRC = path.resolve(`src/musicspot_generator/${FOLDER}/face.png`);
 const GENERATED_DIR = path.resolve('generated');
@@ -84,14 +85,16 @@ async function ensureDirs() {
     await fs.mkdir(GENERATED_DIR, { recursive: true });
 }
 
-async function copyFaceToServers(servers: Server[]): Promise<string> {
+async function copyFaceToServers(servers: Server[]): Promise<string | undefined> {
     const faceFileName = 'face.png';
 
     // Validate face source
     try {
         await fs.access(FACE_SRC);
     } catch {
-        throw new Error(`Face image not found at ${FACE_SRC}`);
+        // If face.png doesn't exist, we don't need to copy anything.
+        // The caller should handle this case.
+        return undefined;
     }
 
     for (const srv of servers) {
@@ -113,14 +116,25 @@ function buildImagePromptRequest(
     character: MusicSpotCharacter,
     scene: MusicSpotScene,
     cut: MusicSpotCut,
-    cameraIntent: string
+    cameraIntent: string,
+    hasFaceImage: boolean // Added parameter
 ): string {
+    let promptInstructions = `
+Write "imagePrompt" in around 110–140 words to generate a still portrait image (720x1280 vertical).
+`;
+
+    if (hasFaceImage) {
+        promptInstructions += `Keep a consistent character identity using the provided face image (identity preservation), but do not mention any camera brand/model.
+`;
+    } else {
+        promptInstructions += `Do not use any face image for identity preservation.
+`;
+    }
+
     return `
 Return exactly one JSON object, nothing else: { "imagePrompt": "Cinematic realistic photo, (camera framing),(character),(pose),(time),(location),(outfit),(action),(lighting)" }.
 
-Write "imagePrompt" in around 110–140 words to generate a still portrait image (720x1280 vertical).
+${promptInstructions}
-Keep a consistent character identity using the provided face image (identity preservation), but do not mention any camera brand/model.
-
 Describe clearly and concretely:
 - Character: ${character.bodyType}; hair: ${character.hairStyle}
 - Camera framing/composition intention: ${cameraIntent}
@@ -178,6 +192,16 @@ async function main() {
     try {
         await ensureDirs();
 
+        // Check if face.png exists
+        let faceImageExists = false;
+        try {
+            await fs.access(FACE_SRC);
+            faceImageExists = true;
+            logger.info(`Face image found at ${FACE_SRC}`);
+        } catch {
+            logger.warn(`Face image not found at ${FACE_SRC}. Images will be generated without face conditioning.`);
+        }
+
         // Load scenes.json
         const configRaw = await fs.readFile(path.resolve(`src/musicspot_generator/${FOLDER}/scenes.json`), 'utf-8');
         const cfg: MusicSpotConfig = JSON.parse(configRaw);
@@ -187,8 +211,13 @@ async function main() {
             return;
         }
 
-        // Ensure face.png in each server's input
+        // Ensure face.png in each server's input, only if it exists
-        const faceFileName = await copyFaceToServers(servers);
+        let faceFileName: string | undefined = undefined;
+        if (faceImageExists) {
+            faceFileName = await copyFaceToServers(servers);
+        } else {
+            logger.info('Skipping copyFaceToServers as face image does not exist.');
+        }
 
         // Generate images only (no video here). Intended to be run first.
         let imageTaskIndex = 0;
@@ -218,7 +247,8 @@ async function main() {
 
                     // 1) Generate image prompt for this camera
                     logger.info(`Scene ${scene.sceneId} - Cut ${cut.cutId} - Cam${variantIndex}: generating image prompt...`);
-                    const imgPromptReq = buildImagePromptRequest(cfg.character, scene, cut, cameraIntent);
+                    // Pass faceImageExists to buildImagePromptRequest
+                    const imgPromptReq = buildImagePromptRequest(cfg.character, scene, cut, cameraIntent, faceImageExists);
                     let imagePrompt: string;
                     try {
                         imagePrompt = await getImagePrompt(imgPromptReq);
@@ -229,18 +259,20 @@ async function main() {
                         continue;
                     }
 
-                    // 2) Generate one image using face conditioning for this specific camera
+                    // 2) Generate one image using the new generateImage function
                     const serverForImage = pickServer(servers, imageTaskIndex++);
                     logger.info(`Generating image (${imgFileName}) on ${serverForImage.name}...`);
 
                     try {
-                        // Use only the face file name for the workflow image input (Comfy expects it in its input dir)
+                        // Use the generic generateImage function from image-generator.ts
-                        const finalImagePath = await generateFaceImage(
+                        // The faceFileName is no longer passed as an argument, but its existence
-                            `Realistic photo, ultra detailed, high contrast, ${imagePrompt}`,
+                        // influenced the imagePrompt.
-                            faceFileName,
+                        const finalImagePath = await generateImage(
+                            imagePrompt, // The prompt now contains face instructions if faceImageExists is true
                             imgFileName,
                             serverForImage.baseUrl!,
                             serverForImage.outputDir!,
+                            'flux', // Use default imageModel ('qwen')
                             DEFAULT_SIZE
                         );
                         logger.info(`Image generated: ${finalImagePath}`);

src/musicspot_generator/videos.ts

@@ -42,8 +42,8 @@ interface Server {
     name: string;
 }
 
-const DEFAULT_SIZE: Size = { width: 720, height: 1280 };
+const DEFAULT_SIZE: Size = { width: 1280, height: 720 };
-const FOLDER = process.argv[2] || process.env.MUSICSPOT_FOLDER || 'zagreb';
+const FOLDER = process.argv[2] || process.env.MUSICSPOT_FOLDER || 'infinitydance';
 const FOLDER_SAFE = FOLDER.replace(/[/\\?%*:|"<>]/g, '_');
 const GENERATED_DIR = path.resolve('generated');
 
@@ -234,7 +234,7 @@ async function main() {
                             serverForVideo.baseUrl!,
                             serverForVideo.outputDir!,
                             DEFAULT_SIZE,
-                            false,
+                            true,
                             true
                         );