Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →PPT 路演视频全流程生成器，支持品牌风格学习、智能配音、音效音乐、字幕和一键视频合成。可一次性生成 15-100 页风格统一的完整路演视频。
.claude/skills/anbeime-ppt-roadshow-generator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 484% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 97% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 176% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 37% | 0% |
moviepy>=1.0.3 pillow>=9.0.0 pydub>=0.25.1 requests>=2.28.0
bash # Ubuntu/Debian sudo apt-get install ffmpeg
# macOS brew install ffmpeg
web-design-analyzer 导出的 brand_style.json，直接导入scripts/style_learner.py --load-json ./brand_style.json 加载配置scripts/audio_processor.py：bash python scripts/audio_processor.py \ --script ./roadshow_script.txt \ --style-brand ./brand_style.json \ --output ./audio/
scripts/subtitle_generator.py：bash python scripts/subtitle_generator.py \ --script ./roadshow_script.txt \ --style-brand ./brand_style.json \ --output ./subtitles.srt
scripts/roadshow_composer.py：bash python scripts/roadshow_composer.py \ --images ./images/ \ --audio ./audio/ \ --subtitles ./subtitles.srt \ --style-brand ./brand_style.json \ --output ./roadshow_video.mp4
当用户希望保持品牌风格一致时：
convert_to_roadshow_style.py）skill_credentials 工具配置| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-22 | fail→fail | 14,582 | 19,713 | +35% | 1 | 1 | 0% | 2,166 | 4,974 | +130% | 0 | 0 | — |
case-15 | pass→pass | 12,736 | 11,811 | -7% | 1 | 1 | 0% | 2,210 | 5,009 | +127% | 0 | 0 | — |
case-21 | fail→fail | 13,446 | 12,998 | -3% | 1 | 1 | 0% | 2,815 | 5,261 | +87% | 0 | 0 | — |
case-01 | fail→pass | 5,268 | 12,022 | +128% | 1 | 1 | 0% | 880 | 5,140 | +484% | 0 | 0 | — |
case-02 | fail→fail | 7,856 | 14,072 | +79% | 1 | 1 | 0% | 1,397 | 5,394 | +286% | 0 | 0 | — |
case-03 | fail→pass | 10,080 | 6,435 | -36% | 1 | 1 | 0% | 1,854 | 3,655 | +97% | 0 | 0 | — |
case-04 | fail→pass | 15,621 | 4,759 | -70% | 1 | 1 | 0% | 3,163 | 3,662 | +16% | 0 | 0 | — |
case-05 | fail→pass | 18,467 | 2,077 | -89% | 1 | 1 | 0% | 1,083 | 2,989 | +176% | 0 | 0 | — |
case-06 | fail→pass | 12,780 | 2,525 | -80% | 1 | 1 | 0% | 2,313 | 3,158 | +37% | 0 | 0 | — |
case-07 | pass→pass | 5,526 | 2,852 | -48% | 1 | 1 | 0% | 1,039 | 3,081 | +197% | 0 | 0 | — |
case-08 | pass→pass | 21,813 | 3,799 | -83% | 1 | 1 | 0% | 1,528 | 3,166 | +107% | 0 | 0 | — |
case-09 | fail→pass | 6,665 | 2,396 | -64% | 1 | 1 | 0% | 991 | 3,077 | +210% | 0 | 0 | — |
case-10 | fail→pass | 9,962 | 6,009 | -40% | 1 | 1 | 0% | 1,646 | 3,508 | +113% | 0 | 0 | — |
case-11 | fail→fail | 8,552 | 5,131 | -40% | 1 | 1 | 0% | 1,559 | 3,784 | +143% | 0 | 0 | — |
case-12 | fail→pass | 5,894 | 5,409 | -8% | 1 | 1 | 0% | 1,226 | 3,744 | +205% | 0 | 0 | — |
case-13 | pass→pass | 4,772 | 3,067 | -36% | 1 | 1 | 0% | 788 | 3,198 | +306% | 0 | 0 | — |
case-14 | fail→pass | 2,500 | 2,205 | -12% | 1 | 1 | 0% | 397 | 3,022 | +661% | 0 | 0 | — |
case-16 | fail→pass | 10,931 | 9,282 | -15% | 1 | 1 | 0% | 1,939 | 4,531 | +134% | 0 | 0 | — |
case-17 | fail→pass | 7,238 | 2,122 | -71% | 1 | 1 | 0% | 1,324 | 3,036 | +129% | 0 | 0 | — |
case-18 | fail→pass | 7,342 | 4,301 | -41% | 1 | 1 | 0% | 1,019 | 3,455 | +239% | 0 | 0 | — |
case-19 | fail→pass | 8,135 | 2,684 | -67% | 1 | 1 | 0% | 1,202 | 3,100 | +158% | 0 | 0 | — |
case-20 | fail→fail | 14,073 | 14,540 | +3% | 1 | 1 | 0% | 2,152 | 5,234 | +143% | 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, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +59 percentage points is the difference between those two pass rates over the 21 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.