Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →完整的视频创作套件，支持原创创作、视频二创、视频分析三种模式，集成Coze Bot API、Edge-TTS、Suno API，涵盖多智能体协同、素材生成、视频合成全流程
.claude/skills/anbeime-video-creation-suite/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 91% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 389% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 223% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 192% | 0% |
opencv-python>=4.8.0
pillow>=10.0.0
moviepy>=1.0.3
numpy>=1.24.0
requests>=2.28.0
edge-tts>=6.1.0COZE_BOT_ID, COZE_API_KEYpip install edge-ttsSUNO_API_KEY智能体根据用户输入自动选择模式：
scripts/voice_generator.py 生成角色配音--input <旁白数据JSON> --output ./output/audio/voicescripts/sound_generator.py --type sound 生成音效--input <音效配置JSON> --output ./output/audio/sound_effectsscripts/sound_generator.py --type music 生成背景音乐--input <音乐配置JSON> --output ./output/audio/background_musicscripts/video_compositor.py 合成最终视频--images <图片目录> --audio <音频目录> --subtitles <字幕文件> --output <输出路径>scripts/quality_checker.py 检测技术指标scripts/video_frame_extractor.py 提取关键帧--input <原视频路径> --output ./output/frames --interval 2./output/frames/scripts/coze_bot_client.py 分析关键帧./output/analysis.jsonscripts/image_generator.py 生成图片./output/images/scripts/voice_generator.py 合成配音（基于Edge-TTS）--input <旁白脚本JSON> --output ./output/voice./output/voice/scripts/sound_generator.py --type both 生成音效和背景音乐--input <完整配置JSON> --output ./output/audio./output/audio/sound_effects/, 背景音乐到 ./output/audio/background_music/scripts/subtitle_generator.py 生成字幕文件--input <字幕数据JSON> --output ./output/subtitles./output/subtitles/scripts/video_compositor.py 合成视频--images ./output/images --audio ./output/audio --voice ./output/voice --subtitles ./output/subtitles --output ./output/final.mp4./output/final.mp4scripts/file_server.py 启动HTTP服务器--port 8080 --directory ./outputhttp://localhost:8080/final.mp4scripts/video_frame_extractor.py 提取关键帧--input <视频路径> --output <输出目录> --interval <间隔秒数>scripts/coze_bot_client.py 分析每帧内容--message "<分析提示>" --image_path <图片路径>用户: "帮我创作一个《三体》威慑纪元的8分钟视频"
智能体执行:
1. 自动选择模式一: 《三体》IP创作
2. 按照前期筹备 → 视觉制作 → 音频与剪辑 → 质量管控流程
3. 调用相应脚本生成配音、音效、背景音乐
4. 合成最终视频用户: "这个视频帮我二创一下，换个风格"
智能体执行:
1. 自动选择模式二: 视频二创
2. 提取原视频关键帧
3. 视觉分析，提取风格特征
4. 生成新素材（图片、配音、音效、音乐）
5. 合成新视频用户: "分析这个视频，提取分镜和提示词"
智能体执行:
1. 自动选择模式三: 视频分析
2. 抽取视频关键帧
3. 分析每帧内容
4. 输出结构化分析结果| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,784 | 17,584 | -23% | 1 | 1 | 0% | 3,462 | 6,535 | +89% | 0 | 0 | — |
case-02 | fail→pass | 15,092 | 8,957 | -41% | 1 | 1 | 0% | 2,627 | 5,017 | +91% | 0 | 0 | — |
case-03 | fail→fail | 22,090 | 8,806 | -60% | 1 | 1 | 0% | 4,251 | 4,896 | +15% | 0 | 0 | — |
case-04 | fail→pass | 3,926 | 3,270 | -17% | 1 | 1 | 0% | 820 | 4,011 | +389% | 0 | 0 | — |
case-19 | pass→pass | 6,419 | 2,927 | -54% | 1 | 1 | 0% | 1,145 | 3,906 | +241% | 0 | 0 | — |
case-05 | fail→pass | 9,406 | 2,308 | -75% | 1 | 1 | 0% | 2,126 | 3,807 | +79% | 0 | 0 | — |
case-06 | fail→pass | 5,877 | 2,207 | -62% | 1 | 1 | 0% | 1,181 | 3,816 | +223% | 0 | 0 | — |
case-07 | fail→pass | 6,421 | 2,521 | -61% | 1 | 1 | 0% | 1,316 | 3,843 | +192% | 0 | 0 | — |
case-08 | fail→pass | 10,175 | 2,635 | -74% | 1 | 1 | 0% | 2,357 | 3,867 | +64% | 0 | 0 | — |
case-09 | fail→pass | 9,509 | 2,412 | -75% | 1 | 1 | 0% | 2,156 | 3,833 | +78% | 0 | 0 | — |
case-10 | fail→pass | 3,005 | 1,920 | -36% | 1 | 1 | 0% | 601 | 3,734 | +521% | 0 | 0 | — |
case-11 | fail→pass | 10,370 | 2,561 | -75% | 1 | 1 | 0% | 2,160 | 3,866 | +79% | 0 | 0 | — |
case-12 | pass→pass | 10,339 | 2,733 | -74% | 1 | 1 | 0% | 2,054 | 3,861 | +88% | 0 | 0 | — |
case-13 | pass→pass | 9,302 | 2,741 | -71% | 1 | 1 | 0% | 1,756 | 3,628 | +107% | 0 | 0 | — |
case-14 | fail→pass | 10,209 | 2,201 | -78% | 1 | 1 | 0% | 1,768 | 3,772 | +113% | 0 | 0 | — |
case-15 | fail→pass | 6,480 | 1,366 | -79% | 1 | 1 | 0% | 1,154 | 3,552 | +208% | 0 | 0 | — |
case-16 | pass→pass | 8,278 | 5,962 | -28% | 1 | 1 | 0% | 1,498 | 4,517 | +202% | 0 | 0 | — |
case-17 | fail→pass | 10,228 | 2,837 | -72% | 1 | 1 | 0% | 1,539 | 3,869 | +151% | 0 | 0 | — |
case-18 | fail→pass | 7,208 | 1,319 | -82% | 1 | 1 | 0% | 1,246 | 3,533 | +184% | 0 | 0 | — |
case-20 | fail→fail | 20,647 | 20,218 | -2% | 1 | 1 | 0% | 3,740 | 7,368 | +97% | 0 | 0 | — |
case-21 | fail→pass | 16,331 | 10,392 | -36% | 1 | 1 | 0% | 3,350 | 5,546 | +66% | 0 | 0 | — |
case-22 | fail→fail | 15,145 | 11,437 | -24% | 1 | 1 | 0% | 3,062 | 5,567 | +82% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +64 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.