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

src/comfyworkflows/edit_image_2_qwen.json

@@ -57,7 +57,7 @@
     },
     "7": {
         "inputs": {
-            "seed": 838097333311955,
+            "seed": 920668017513581,
             "steps": 8,
             "cfg": 1,
             "sampler_name": "euler",
@@ -101,14 +101,56 @@
             "title": "VAE Decode"
         }
     },
+    "9": {
+        "inputs": {
+            "font_file": "Alibaba-PuHuiTi-Heavy.ttf",
+            "font_size": 40,
+            "border": 32,
+            "color_theme": "light",
+            "reel_1": [
+                "10",
+                0
+            ]
+        },
+        "class_type": "LayerUtility: ImageReelComposit",
+        "_meta": {
+            "title": "LayerUtility: Image Reel Composit"
+        }
+    },
+    "10": {
+        "inputs": {
+            "image1_text": "Original image",
+            "image2_text": "Reference",
+            "image3_text": "Result",
+            "image4_text": "image4",
+            "reel_height": 512,
+            "border": 32,
+            "image1": [
+                "11",
+                1
+            ],
+            "image2": [
+                "11",
+                2
+            ],
+            "image3": [
+                "8",
+                0
+            ]
+        },
+        "class_type": "LayerUtility: ImageReel",
+        "_meta": {
+            "title": "LayerUtility: Image Reel"
+        }
+    },
     "11": {
         "inputs": {
             "prompt": [
                 "21",
                 0
             ],
-            "enable_resize": true,
+            "enable_resize": false,
-            "enable_vl_resize": true,
+            "enable_vl_resize": false,
             "upscale_method": "lanczos",
             "crop": "disabled",
             "instruction": "<|im_start|>system\nDescribe the key features of the input image (color, shape, size, texture, objects, background), then explain how the user's text instruction should alter or modify the image. Generate a new image that meets the user's requirements while maintaining consistency with the original input where appropriate.<|im_end|>\n<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n",
@@ -121,11 +163,11 @@
                 0
             ],
             "image1": [
-                "24",
+                "23",
                 0
             ],
             "image2": [
-                "15",
+                "24",
                 0
             ]
         },
@@ -134,9 +176,18 @@
             "title": "TextEncodeQwenImageEditPlus 小志Jason(xiaozhijason)"
         }
     },
+    "14": {
+        "inputs": {
+            "image": "f81662775bd0e7950e4794933ef4b3d973fbb9c2db397c8b46809797954d0074.png"
+        },
+        "class_type": "LoadImage",
+        "_meta": {
+            "title": "Load Image"
+        }
+    },
     "15": {
         "inputs": {
-            "image": "ComfyUI_00067_.png"
+            "image": "monster_554.png"
         },
         "class_type": "LoadImage",
         "_meta": {
@@ -145,7 +196,7 @@
     },
     "20": {
         "inputs": {
-            "filename_prefix": "qwenedit",
+            "filename_prefix": "combined",
             "images": [
                 "8",
                 0
@@ -158,49 +209,94 @@
     },
     "21": {
         "inputs": {
-            "value": "change camera angle to closeup face from image1, change background to light gray with faing gradient, change face angle to look at directry look at camera"
+            "value": "只提取图2中的怪物，怪物站在图1的女生身后，使用图1的背景，并调整怪物的光线以符合图1。\n\n\n\n\n\n\n"
         },
         "class_type": "PrimitiveStringMultiline",
         "_meta": {
             "title": "String (Multiline)"
         }
     },
+    "22": {
+        "inputs": {
+            "filename_prefix": "ComfyUI",
+            "images": [
+                "9",
+                0
+            ]
+        },
+        "class_type": "SaveImage",
+        "_meta": {
+            "title": "Save Image"
+        }
+    },
+    "23": {
+        "inputs": {
+            "width": [
+                "25",
+                0
+            ],
+            "height": [
+                "26",
+                0
+            ],
+            "upscale_method": "nearest-exact",
+            "keep_proportion": "stretch",
+            "pad_color": "0, 0, 0",
+            "crop_position": "center",
+            "divisible_by": 2,
+            "device": "cpu",
+            "image": [
+                "14",
+                0
+            ]
+        },
+        "class_type": "ImageResizeKJv2",
+        "_meta": {
+            "title": "Resize Image v2"
+        }
+    },
     "24": {
         "inputs": {
-            "measurement": "pixels",
+            "width": [
-            "width": 720,
+                "25",
-            "height": 1280,
+                0
-            "fit": "contain",
+            ],
-            "method": "nearest-exact",
+            "height": [
+                "26",
+                0
+            ],
+            "upscale_method": "nearest-exact",
+            "keep_proportion": "stretch",
+            "pad_color": "0, 0, 0",
+            "crop_position": "center",
+            "divisible_by": 2,
+            "device": "cpu",
             "image": [
-                "64",
+                "15",
                 0
             ]
         },
-        "class_type": "Image Resize (rgthree)",
+        "class_type": "ImageResizeKJv2",
         "_meta": {
-            "title": "Image Resize (rgthree)"
+            "title": "Resize Image v2"
         }
     },
-    "64": {
+    "25": {
         "inputs": {
-            "image": "1337074888177434_1758776251440_2.png"
+            "Number": "1280"
         },
-        "class_type": "LoadImage",
+        "class_type": "Int",
         "_meta": {
-            "title": "Load Image"
+            "title": "width"
         }
     },
-    "65": {
+    "26": {
         "inputs": {
-            "images": [
+            "Number": "720"
-                "24",
-                0
-            ]
         },
-        "class_type": "PreviewImage",
+        "class_type": "Int",
         "_meta": {
-            "title": "Preview Image"
+            "title": "height"
         }
     }
 }

src/lib/image-converter.ts

@@ -81,9 +81,9 @@ async function convertImageWithFile(
 
     workflow = JSON.parse(await fs.readFile('src/comfyworkflows/edit_image_2_qwen.json', 'utf-8'));
     workflow['21']['inputs']['value'] = prompt;
-    workflow['24']['inputs']['width'] = size.width;
+    workflow['25']['inputs']['width'] = size.width;
-    workflow['24']['inputs']['height'] = size.height;
+    workflow['26']['inputs']['height'] = size.height;
-    workflow['64']['inputs']['image'] = baseFileName;
+    workflow['14']['inputs']['image'] = baseFileName;
     workflow['15']['inputs']['image'] = secondFileName;
 
     const response = await axios.post(`${COMFY_BASE_URL}/prompt`, { prompt: workflow });
@@ -97,7 +97,7 @@ async function convertImageWithFile(
     } while (!history || Object.keys(history.outputs).length === 0);
 
     const files = await fs.readdir(COMFY_OUTPUT_DIR!);
-    const generatedFiles = files.filter(file => file.startsWith('qwenedit'));
+    const generatedFiles = files.filter(file => file.startsWith('combined'));
 
     const fileStats = await Promise.all(
         generatedFiles.map(async (file) => {

src/musicspot_generator/v2/combine_image.ts

@@ -0,0 +1,78 @@
+import * as fs from 'fs';
+import * as path from 'path';
+import { convertImageWithFile } from '../../lib/image-converter';
+import dotenv from 'dotenv';
+
+dotenv.config();
+
+const girlDir = path.join(__dirname, '../../../input/girl');
+const monsterDir = path.join(__dirname, '../../../input/monster');
+const outputDir = path.join(__dirname, '../../../generated');
+const prompt = "只提取图1中的女生，把她放在图2的怪物之间。";
+
+if (!fs.existsSync(outputDir)) {
+    fs.mkdirSync(outputDir, { recursive: true });
+}
+
+const comfyBaseUrl = process.env.SERVER2_COMFY_BASE_URL;
+const comfyOutputDir = process.env.SERVER2_COMFY_OUTPUT_DIR;
+
+if (!comfyBaseUrl || !comfyOutputDir) {
+    console.error("Please define SERVER1_COMFY_BASE_URL and SERVER1_COMFY_OUTPUT_DIR in your .env file");
+    process.exit(1);
+}
+
+const comfyInputDir = comfyOutputDir.replace("output", "input");
+if (!fs.existsSync(comfyInputDir)) {
+    fs.mkdirSync(comfyInputDir, { recursive: true });
+}
+
+async function combineImages() {
+    while (true) {
+        try {
+            const girlImages = fs.readdirSync(girlDir).filter(file => /\.(jpg|jpeg|png)$/i.test(file));
+            const monsterImages = fs.readdirSync(monsterDir).filter(file => /\.(jpg|jpeg|png)$/i.test(file));
+
+            if (girlImages.length === 0 || monsterImages.length === 0) {
+                console.log('Input directories are empty. Waiting...');
+                await new Promise(resolve => setTimeout(resolve, 5000));
+                continue;
+            }
+
+            const randomGirlImage = girlImages[Math.floor(Math.random() * girlImages.length)];
+            const randomMonsterImage = monsterImages[Math.floor(Math.random() * monsterImages.length)];
+
+            const image1Path = path.join(girlDir, randomGirlImage);
+            const image2Path = path.join(monsterDir, randomMonsterImage);
+
+            // Copy files to comfy input directory
+            const destImage1Path = path.join(comfyInputDir, randomGirlImage);
+            const destImage2Path = path.join(comfyInputDir, randomMonsterImage);
+            fs.copyFileSync(image1Path, destImage1Path);
+            fs.copyFileSync(image2Path, destImage2Path);
+
+            console.log(`Combining ${randomGirlImage} and ${randomMonsterImage}`);
+
+            const generatedFilePath = await convertImageWithFile(prompt, randomGirlImage, randomMonsterImage, comfyBaseUrl!, comfyOutputDir!);
+
+            if (generatedFilePath && fs.existsSync(generatedFilePath)) {
+                const timestamp = new Date().getTime();
+                const newFileName = `combined_${timestamp}.png`;
+                const newFilePath = path.join(outputDir, newFileName);
+
+                fs.renameSync(generatedFilePath, newFilePath);
+                console.log(`Renamed generated file to ${newFilePath}`);
+            } else {
+                console.log("Failed to generate or find the image file.");
+            }
+
+        } catch (error) {
+            console.error('An error occurred:', error);
+        }
+
+        // Wait for a bit before the next iteration
+        await new Promise(resolve => setTimeout(resolve, 5000));
+    }
+}
+
+combineImages();

src/musicspot_generator/v2/generate_photo.ts

@@ -52,7 +52,7 @@ async function generatePhotos() {
 
         try {
             await generateImage(
-                `realistic photo of woman, wavy long brown hair, fullbody shot, ${scene.imagePrompt.location},${scene.imagePrompt.angle},${scene.imagePrompt.lighting},${scene.imagePrompt.outfit}`,
+                `Scary realistic photo, ${scene.imagePrompt.location},${scene.imagePrompt.angle},${scene.imagePrompt.lighting},${scene.imagePrompt.outfit}`,
                 faceFilePath,
                 scene.baseImagePath,
                 imgFileName,

src/musicspot_generator/v2/generate_scenes.ts

@@ -7,7 +7,7 @@ const promptInstructions = `
     Video prompt: No slowmotion, Be creative and generate gengle action scene.
     `;
 6
-const inputDir = path.resolve(process.cwd(), 'input');
+const inputDir = path.resolve(process.cwd(), 'input/static');
 const outputFilePath = path.resolve(process.cwd(), 'src/musicspot_generator/v2/scenes.json');
 
 interface Scene {

src/musicspot_generator/v2/generate_video_from_static_image.ts

@@ -7,7 +7,7 @@ import dotenv from 'dotenv';
 
 dotenv.config();
 
-const inputFolderPath = path.join(__dirname, '..', '..', '..', 'input');
+const inputFolderPath = path.join(__dirname, '..', '..', '..', 'input/static');
 const generatedFolderPath = path.join(__dirname, '..', '..', '..', 'generated');
 
 async function processImages() {

src/musicspot_generator/v2/photo_download.ts

@@ -13,8 +13,8 @@ const PINS_TO_COLLECT = 5;
 
 // Hard-coded user prompt
 const HARDCODED_USER_PROMPT = process.env.HARDCODED_USER_PROMPT || `
-    Generate 20 keywords for photos of a girl in scary scene for music video for haloween. All keywords should containe girl in a scene.
+    Generate 20 keywords for photos of a ghotst or monster from all over the world. "Cute Japanese yokai" is mandatory, also add "Realistic photo cute" keyword to all genearated keywords first.
-    Example output : ["a girl in grave yard,"a girl in scary forest","",... and 20 items in array]
+    Example output : ["Cute Japanese yokai","Realistic photo Cute ghost","Realistic photo cute monster","Realistic photo cute haloween monster","Realistic photo cute haloween ghost"... and 20 items in array]
     `;
 
 async function getPinUrlsFromPinterest(keyword: string, scrollCount = SCROLL_SEARCH, limit = PINS_TO_COLLECT): Promise<string[]> {