cerevox 2.19.0 → 2.20.1

package/dist/mcp/servers/zerocut.js

@@ -1003,7 +1003,11 @@ server.registerTool('generate-image', {
         prompt: zod_1.z
             .string()
             .describe('The prompt to generate. 一般要严格对应 story_board 中当前场景的 start_frame 或 end_frame 中的字段描述'),
-        sceneIndex: zod_1.z.number().min(1).describe('场景索引，从1开始的下标'),
+        sceneIndex: zod_1.z
+            .number()
+            .min(1)
+            .optional()
+            .describe('场景索引，从1开始的下标，如果非场景对应素材，则可不传，场景素材必传'),
         storyBoardFile: zod_1.z
             .string()
             .optional()
@@ -1088,16 +1092,16 @@ server.registerTool('generate-image', {
     try {
         // 验证session状态
         const currentSession = await validateSession('generate-image');
+        const storyBoardPath = (0, node_path_1.resolve)(process.env.ZEROCUT_PROJECT_CWD || process.cwd(), projectLocalDir, storyBoardFile);
         // 检查 storyboard 标志
-        if (!checkStoryboardFlag) {
+        if (!checkStoryboardFlag && (0, node_fs_1.existsSync)(storyBoardPath)) {
             checkStoryboardFlag = true;
             return createErrorResponse('必须先审查生成的 story_board.json 内容，确保每个场景中的stage_atmosphere内容按照规则被正确融合到start_frame和video_prompt中，不得遗漏，检查完成后先汇报，如果有问题，应当先修改 story_board.json 内容，然后再调用 generate-image 生成图片。注意修改 story_board 内容时，仅修改相应字段的字符串值，不要破坏JSON格式！', 'generate-image');
         }
         const validatedFileName = validateFileName(saveToFileName);
         // 校验 prompt 与 story_board.json 中场景设定的一致性
-        if (!skipConsistencyCheck) {
+        if (sceneIndex && !skipConsistencyCheck) {
             try {
-                const storyBoardPath = (0, node_path_1.resolve)(process.env.ZEROCUT_PROJECT_CWD || process.cwd(), projectLocalDir, storyBoardFile);
                 if ((0, node_fs_1.existsSync)(storyBoardPath)) {
                     const storyBoardContent = await (0, promises_1.readFile)(storyBoardPath, 'utf8');
                     // 检查 storyBoard JSON 语法合法性
@@ -1343,249 +1347,6 @@ ${processedPrompt}`.trim();
         return createErrorResponse(error, 'generate-image');
     }
 });
-server.registerTool('generate-image-series', {
-    title: 'Generate Image Series',
-    description: 'Generate a series of images based on prompts.',
-    inputSchema: {
-        prompts: zod_1.z
-            .union([zod_1.z.string(), zod_1.z.array(zod_1.z.string())])
-            .describe('The prompts to generate images. Can be a single string or array of strings.'),
-        size: zod_1.z
-            .enum([
-            '1024x1024',
-            '864x1152',
-            '1152x864',
-            '1280x720',
-            '720x1280',
-            '832x1248',
-            '1248x832',
-            '1512x648',
-            // 2K
-            '2048x2048',
-            '1728x2304',
-            '2304x1728',
-            '2560x1440',
-            '1440x2560',
-            '1664x2496',
-            '2496x1664',
-            '3024x1456',
-            // 4K
-            '4096x4096',
-            '3072x4096',
-            '4096x3072',
-            '4096x2304',
-            '2304x4096',
-            '2731x4096',
-            '4096x2731',
-            '4096x1968',
-        ])
-            .default('720x1280')
-            .describe('The size of the image.'),
-        watermark: zod_1.z
-            .boolean()
-            .optional()
-            .default(false)
-            .describe('Whether to add watermark to the images.'),
-        max_count: zod_1.z
-            .number()
-            .min(1)
-            .max(15)
-            .optional()
-            .default(15)
-            .describe('Maximum number of images to generate (1-15).'),
-        referenceImages: zod_1.z
-            .array(zod_1.z.object({
-            image: zod_1.z.string().describe('Local image file path'),
-            type: zod_1.z
-                .enum(['character', 'object', 'background'])
-                .describe('Type of the reference image. 必须传，如果是参考角色三视图，传character，如果是参考背景图，传background，否则传object'),
-            name: zod_1.z.string().describe('Name for this reference image'),
-            isTurnaround: zod_1.z
-                .boolean()
-                .describe('Whether this is a turnaround image.如果是三视图，这个参数务必传true'),
-        }))
-            .optional()
-            .describe(`Array of reference images with character or object names.如果stage_atmosphere中有角色apply_reference_image，那么必须要传这个参数生成分镜图片
-
-传参示例
-\`\`\`
-{
-  "image": "latiao.jpeg",
-  "type": "object",
-  "name": "卫龙辣条",
-}
-\`\`\`
-          `),
-        saveToFileNames: zod_1.z
-            .array(zod_1.z.string())
-            .describe('Array of filenames to save the generated images.'),
-    },
-}, async ({ prompts, size, watermark, max_count, referenceImages, saveToFileNames }, context) => {
-    try {
-        const currentSession = await validateSession('generate-image-series');
-        // Validate file names
-        const validatedFileNames = saveToFileNames.map(fileName => validateFileName(fileName));
-        if (typeof prompts === 'string' && prompts.startsWith('[')) {
-            try {
-                prompts = JSON.parse(prompts);
-            }
-            catch (ex) {
-                // 解析失败，保持原始字符串
-            }
-        }
-        console.log(`Generating image series with ${Array.isArray(prompts) ? prompts.length : 1} prompts...`);
-        // 处理参考图片
-        let imageBase64Array;
-        let refPrefix = '';
-        if (referenceImages && referenceImages.length > 0) {
-            imageBase64Array = [];
-            const objectPrefix = [];
-            let index = 0;
-            let hasTurnaround = false;
-            for (const refImage of referenceImages) {
-                const imagePath = (0, node_path_1.dirname)(refImage.image) !== '.'
-                    ? refImage.image
-                    : `./materials/${refImage.image}`;
-                // 需要得到当前项目的绝对路径
-                const imageFilePath = (0, node_path_1.resolve)(process.env.ZEROCUT_PROJECT_CWD || process.cwd(), projectLocalDir, imagePath);
-                try {
-                    // 直接读取本地文件
-                    if (!(0, node_fs_1.existsSync)(imageFilePath)) {
-                        return createErrorResponse(`Reference image not found: ${imageFilePath}`, 'generate-image-series');
-                    }
-                    const imageBuffer = await (0, promises_1.readFile)(imageFilePath);
-                    const fileName = (0, node_path_1.basename)(imagePath);
-                    const mimeType = fileName.toLowerCase().endsWith('.png')
-                        ? 'image/png'
-                        : fileName.toLowerCase().endsWith('.jpg') ||
-                            fileName.toLowerCase().endsWith('.jpeg')
-                            ? 'image/jpeg'
-                            : 'image/png';
-                    const base64String = `data:${mimeType};base64,${imageBuffer.toString('base64')}`;
-                    imageBase64Array.push(base64String);
-                    console.log(`Loaded reference image: ${imagePath} for character: ${refImage.name}`);
-                    if (refImage.type === 'character') {
-                        if (refImage.isTurnaround) {
-                            objectPrefix.push(`[参考图${++index}]是名为"${refImage.name}"的人物角色三视图`);
-                            hasTurnaround = true;
-                        }
-                        else {
-                            objectPrefix.push(`[参考图${++index}]是名为"${refImage.name}"的人物角色形象`);
-                        }
-                    }
-                    else if (refImage.type === 'object') {
-                        if (refImage.isTurnaround) {
-                            objectPrefix.push(`[参考图${++index}]是名为"${refImage.name}"的物件三视图`);
-                            hasTurnaround = true;
-                        }
-                        else {
-                            objectPrefix.push(`[参考图${++index}]是名为"${refImage.name}"的物件`);
-                        }
-                    }
-                    else if (refImage.type === 'background') {
-                        objectPrefix.push(`[参考图${++index}]是背景图，描述了"${refImage.name}"`);
-                    }
-                }
-                catch (error) {
-                    console.error(`Failed to load reference image ${imageFilePath} for ${refImage.name}:`, error);
-                    return createErrorResponse(`Failed to load reference image ${imageFilePath} for ${refImage.name}: ${error}`, 'generate-image-series');
-                }
-            }
-            const turnaroundMessage = hasTurnaround
-                ? '\n\n⚠️ 注意**三视图**是指**同一个**人或物体的不同视角合成图，三部分都表示同一个人或物体，只能参考其信息画出**一个人或一个物体**，不要画成多个人或多个物体！\n'
-                : '';
-            if (objectPrefix.length > 0) {
-                refPrefix = `${objectPrefix.join('\n')}${turnaroundMessage}`;
-            }
-        }
-        const ai = currentSession.ai;
-        let progress = 0;
-        const res = await ai.generateImageSeries({
-            prompts,
-            refPrefix,
-            size,
-            watermark,
-            image: imageBase64Array,
-            max_count,
-            onProgress: async (metaData) => {
-                // 心跳机制，每分钟调用一次，传递空的 metaData
-                try {
-                    await sendProgress(context, ++progress, undefined, JSON.stringify(metaData));
-                }
-                catch (progressError) {
-                    console.warn('Failed to send progress update:', progressError);
-                }
-            },
-        });
-        if (!res) {
-            throw new Error('Failed to generate image series: no response from AI service');
-        }
-        if (res.urls && Array.isArray(res.urls)) {
-            console.log(`Image series generated successfully (${res.urls.length} images), saving to materials...`);
-            const results = [];
-            const maxImages = Math.min(res.urls.length, validatedFileNames.length);
-            for (let i = 0; i < maxImages; i++) {
-                const url = res.urls[i];
-                const fileName = validatedFileNames[i];
-                try {
-                    const uri = await saveMaterial(currentSession, url, fileName);
-                    results.push({
-                        success: true,
-                        uri,
-                        fileName,
-                        index: i,
-                        timestamp: new Date().toISOString(),
-                    });
-                }
-                catch (error) {
-                    console.error(`Failed to save image ${i + 1}:`, error);
-                    results.push({
-                        success: false,
-                        error: `Failed to save image ${i + 1}: ${error}`,
-                        fileName,
-                        index: i,
-                        timestamp: new Date().toISOString(),
-                    });
-                }
-            }
-            const result = {
-                success: true,
-                prompt: res.prompt,
-                totalGenerated: res.urls.length,
-                totalSaved: results.filter(r => r.success).length,
-                results,
-                timestamp: new Date().toISOString(),
-            };
-            return {
-                content: [
-                    {
-                        type: 'text',
-                        text: JSON.stringify(result),
-                    },
-                ],
-            };
-        }
-        else {
-            console.warn('Image series generation completed but no URLs returned');
-            return {
-                content: [
-                    {
-                        type: 'text',
-                        text: JSON.stringify({
-                            success: false,
-                            error: 'No image URLs returned from AI service',
-                            response: res,
-                            timestamp: new Date().toISOString(),
-                        }),
-                    },
-                ],
-            };
-        }
-    }
-    catch (error) {
-        return createErrorResponse(error, 'generate-image-series');
-    }
-});
 server.registerTool('edit-image', {
     title: 'Edit Image',
     description: 'Edit the image.',
@@ -1696,7 +1457,11 @@ server.registerTool('generate-video', {
         prompt: zod_1.z
             .string()
             .describe('The prompt to generate. 一般要严格对应 story_board 中当前场景的 video_prompt 字段描述'),
-        sceneIndex: zod_1.z.number().min(1).describe('场景索引，从1开始的下标'),
+        sceneIndex: zod_1.z
+            .number()
+            .min(1)
+            .optional()
+            .describe('场景索引，从1开始的下标，如果非场景对应素材，则可不传，场景素材必传'),
         storyBoardFile: zod_1.z
             .string()
             .optional()
@@ -1754,7 +1519,7 @@ server.registerTool('generate-video', {
         // 验证session状态
         const currentSession = await validateSession('generate-video');
         // 校验 prompt 与 story_board.json 中场景设定的一致性以及视频时长与 timeline_analysis.json 中 proposed_video_scenes 的匹配
-        if (!skipConsistencyCheck) {
+        if (sceneIndex && !skipConsistencyCheck) {
             try {
                 const storyBoardPath = (0, node_path_1.resolve)(process.env.ZEROCUT_PROJECT_CWD || process.cwd(), projectLocalDir, storyBoardFile);
                 if ((0, node_fs_1.existsSync)(storyBoardPath)) {
@@ -2453,7 +2218,11 @@ server.registerTool('generate-scene-tts', {
     description: `生成场景配音`,
     inputSchema: {
         text: zod_1.z.string().describe('The text to generate.'),
-        sceneIndex: zod_1.z.number().min(1).describe('场景索引，从1开始的下标'),
+        sceneIndex: zod_1.z
+            .number()
+            .min(1)
+            .optional()
+            .describe('场景索引，从1开始的下标，如果非场景对应素材，则可不传，场景素材必传'),
         storyBoardFile: zod_1.z
             .string()
             .optional()
@@ -2523,7 +2292,7 @@ server.registerTool('generate-scene-tts', {
         const finalSpeed = speed ?? 1;
         volume = volume ?? 1;
         // 校验 text 与 story_board.json 中场景设定的一致性
-        if (!skipConsistencyCheck) {
+        if (sceneIndex && !skipConsistencyCheck) {
             try {
                 const storyBoardPath = (0, node_path_1.resolve)(process.env.ZEROCUT_PROJECT_CWD || process.cwd(), projectLocalDir, storyBoardFile);
                 if ((0, node_fs_1.existsSync)(storyBoardPath)) {
@@ -2697,7 +2466,7 @@ server.registerTool('compile-and-run', {
         // 检查字幕内容匹配标记
         if (!checkStoryboardSubtitlesFlag) {
             checkStoryboardSubtitlesFlag = true;
-            return createErrorResponse('请先检查字幕文字内容是否与 story_board 中各个场景的 script 或 dialog 内容完全匹配，若不匹配，修改 draft_content 使其匹配，返回检查结果，然后再进行合成', 'compile-and-run');
+            return createErrorResponse('请先检查字幕文字内容是否与 story_board 中各个场景的 script 或 dialog 内容完全一致（允许字幕分段，只要最终文本一致就行），若不匹配，修改 draft_content 使其匹配，返回检查结果，然后再进行合成', 'compile-and-run');
         }
         console.log('Starting video compilation and rendering...');
         // 验证terminal可用性
@@ -3476,7 +3245,7 @@ server.registerTool('build-capcat-draft', {
 });
 server.registerTool('generate-video-by-ref', {
     title: 'Generate Video by Reference Images',
-    description: 'Generate video using reference images. Supports sora2 (1 image max), lite and pro (4 images max). Can work without reference images (0 images).',
+    description: 'Generate video using reference images. Supports sora2, veo3.1, veo3.1-pro (1 image max), lite and pro (4 images max), vidu (7 images max). Can work without reference images (0 images).',
     inputSchema: {
         prompt: zod_1.z
             .string()
@@ -3504,7 +3273,7 @@ server.registerTool('generate-video-by-ref', {
             .default(false)
             .describe('Whether to add watermark to the video.'),
         type: zod_1.z
-            .enum(['lite', 'pro', 'sora2', 'vidu'])
+            .enum(['lite', 'sora2', 'veo3.1', 'veo3.1-pro', 'vidu'])
             .optional()
             .default('lite')
             .describe('The model type to use. sora2 allows max 1 reference image, lite and pro allow max 4.'),
@@ -3516,18 +3285,122 @@ server.registerTool('generate-video-by-ref', {
         saveToFileName: zod_1.z
             .string()
             .describe('The filename to save the generated video.'),
+        sceneIndex: zod_1.z
+            .number()
+            .min(1)
+            .optional()
+            .describe('场景索引，从1开始的下标，如果非场景对应素材，则可不传，场景素材必传'),
+        storyBoardFile: zod_1.z
+            .string()
+            .optional()
+            .default('story_board.json')
+            .describe('故事板文件路径'),
+        skipConsistencyCheck: zod_1.z
+            .boolean()
+            .optional()
+            .default(false)
+            .describe('是否跳过一致性检查，默认为false（即默认进行一致性检查）'),
+        skipCheckWithSceneReason: zod_1.z
+            .string()
+            .optional()
+            .describe('跳过校验的理由，如果skipConsistencyCheck设为true，必须要传这个参数'),
+        optimizePrompt: zod_1.z
+            .boolean()
+            .optional()
+            .default(false)
+            .describe('Whether to optimize the prompt.'),
     },
-}, async ({ prompt, referenceImages, duration, size, watermark, type, mute, saveToFileName, }, context) => {
+}, async ({ prompt, referenceImages, duration, size, watermark, type, mute, saveToFileName, sceneIndex, storyBoardFile, skipConsistencyCheck, optimizePrompt, }, context) => {
     try {
         // 验证session状态
         const currentSession = await validateSession('generate-video-by-ref');
+        const storyBoardPath = (0, node_path_1.resolve)(process.env.ZEROCUT_PROJECT_CWD || process.cwd(), projectLocalDir, storyBoardFile);
+        // 检查 storyboard 标志
+        if (!checkStoryboardFlag && (0, node_fs_1.existsSync)(storyBoardPath)) {
+            checkStoryboardFlag = true;
+            return createErrorResponse(`必须先审查生成的 story_board.json 内容，按照如下步骤：
+
+1. 确保每个场景中的stage_atmosphere内容按照规则被正确融合到video_prompt中，不得遗漏
+2. 如有main_characters设定且包含了reference_image，或有reference_objects，需确保video_prompt描述已包含该场景相关main_characters和所有reference_objects中的物品或背景，并确保参考图具体内容已经在video_prompt中有明确描述，如果没有，可忽略。
+3. 如有配音，先自我检查 media_logs 中的查音频时长，确保以匹配音频时长来生成视频
+
+检查完上述问题后先汇报，如果有需要，应当先修改 story_board.json 内容，然后再调用 generate-video-by-ref 生成视频。注意修改 story_board 内容时，仅修改相应字段的字符串值，不要破坏JSON格式！
+
+再次调用 generate-video-by-ref 时，如需要参考图，要确保referenceImages使用正确（main_characters中的reference_image作为参考人物，reference_objects中的image作为参考物品或参考背景）`, 'generate-image');
+        }
+        // 校验 prompt 与 story_board.json 中场景设定的一致性（如果提供了 sceneIndex）
+        if (!skipConsistencyCheck && sceneIndex) {
+            try {
+                if ((0, node_fs_1.existsSync)(storyBoardPath)) {
+                    const storyBoardContent = await (0, promises_1.readFile)(storyBoardPath, 'utf8');
+                    // 检查 storyBoard JSON 语法合法性
+                    let storyBoard;
+                    try {
+                        storyBoard = JSON.parse(storyBoardContent);
+                    }
+                    catch (jsonError) {
+                        return createErrorResponse(`storyBoard 文件 ${storyBoardFile} 存在 JSON 语法错误，请修复后重试。错误详情: ${jsonError instanceof Error ? jsonError.message : String(jsonError)}`, 'generate-video-by-ref');
+                    }
+                    if (storyBoard.scenes && Array.isArray(storyBoard.scenes)) {
+                        const scene = storyBoard.scenes[sceneIndex - 1]; // sceneIndex 从1开始，数组从0开始
+                        if (scene) {
+                            const videoPrompt = scene.video_prompt;
+                            if (videoPrompt && prompt !== videoPrompt) {
+                                return createErrorResponse('视频提示词必须严格遵照story_board的设定，如果用户明确指出不需要遵守，请将skipConsistencyCheck设置为true后再次调用', 'generate-video-by-ref');
+                            }
+                            // 检查 scene.is_continuous 是否为 true
+                            if (scene.is_continuous === true) {
+                                return createErrorResponse('连续镜头应使用首尾帧，请修改连续镜头设置，或将本场景改为首尾帧方式实现', 'generate-video-by-ref');
+                            }
+                        }
+                        else {
+                            console.warn(`Scene index ${sceneIndex} not found in story_board.json`);
+                        }
+                    }
+                }
+                else {
+                    console.warn(`Story board file not found: ${storyBoardPath}`);
+                }
+            }
+            catch (error) {
+                console.error('Failed to validate prompt with story board:', error);
+                // 如果读取或解析 story_board.json 失败，继续执行但记录警告
+            }
+        }
         const validatedFileName = validateFileName(saveToFileName);
+        // 优化提示词（如果启用）
+        const ai = currentSession.ai;
+        if (optimizePrompt) {
+            try {
+                const promptOptimizer = await (0, promises_1.readFile)((0, node_path_1.resolve)(__dirname, './prompts/video-prompt-optimizer.md'), 'utf8');
+                const completion = await ai.getCompletions({
+                    messages: [
+                        {
+                            role: 'system',
+                            content: promptOptimizer,
+                        },
+                        {
+                            role: 'user',
+                            content: prompt.trim(),
+                        },
+                    ],
+                });
+                const optimizedPrompt = completion.choices[0]?.message?.content;
+                if (optimizedPrompt) {
+                    prompt = optimizedPrompt;
+                }
+            }
+            catch (error) {
+                console.error('Failed to optimize prompt:', error);
+            }
+        }
         // 验证参考图数量限制
-        if (type === 'sora2' && referenceImages.length > 1) {
-            return createErrorResponse('sora2 model only supports maximum 1 reference image', 'generate-video-by-ref');
+        if ((type === 'sora2' || type === 'veo3.1' || type === 'veo3.1-pro') &&
+            referenceImages.length > 1) {
+            return createErrorResponse(`${type} model only supports maximum 1 reference image`, 'generate-video-by-ref');
         }
-        if ((type === 'lite' || type === 'pro') && referenceImages.length > 4) {
-            return createErrorResponse('lite and pro models only support maximum 4 reference images', 'generate-video-by-ref');
+        if (type === 'lite' && referenceImages.length > 4) {
+            return createErrorResponse('lite model only support maximum 4 reference images', 'generate-video-by-ref');
         }
         if (type === 'vidu' && referenceImages.length > 7) {
             return createErrorResponse('vidu models only support maximum 7 reference images', 'generate-video-by-ref');
@@ -3593,7 +3466,7 @@ server.registerTool('generate-video-by-ref', {
     }
     catch (error) {
         console.error('Error generating video by reference:', error);
-        return createErrorResponse(error, 'generate-video-by-ref');
+        return createErrorResponse(error.message, 'generate-video-by-ref');
     }
 });
 server.registerTool('run-ffmpeg-command', {