Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →用于录音转文本的音视频转写原子 skill，适用于通用行业文档解析场景。
.claude/skills/aifinlab-audio-video-transcription/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 667% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 73% | 0% |
本 Skill 支持多种音视频文件输入格式，核心数据来源包括：
> 说明：本 Skill 不包含文件采集功能，需要用户提供音视频文件。建议文件质量良好，以便进行准确的转写。
本 Skill 提供全面的音视频转写能力，涵盖多种转写功能：
json{ "file_info": { "filename": "meeting_recording.mp3", "file_size": 51200000, "duration": 3600, "format": "mp3", "language": "zh-CN" }, "transcription": { "full_text": "会议转写完整文本内容...", "segments": [ { "start_time": 0, "end_time": 120, "speaker": "speaker_1", "text": "大家好，今天我们讨论一下项目进展。", "confidence": 0.95 }, { "start_time": 120, "end_time": 240, "speaker": "speaker_2", "text": "项目目前进展顺利，已完成80%的工作。", "confidence": 0.92 } ] }, "speakers": [ { "speaker_id": "speaker_1", "name": "张三", "total_duration": 1800, "segment_count": 15 }, { "speaker_id": "speaker_2", "name": "李四", "total_duration": 1200, "segment_count": 10 } ], "keywords": [ "项目进展", "完成度", "下一步计划" ], "summary": "会议主要讨论了项目进展情况，目前已完成80%的工作，下一步将进行测试和验收。" }
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→pass | 17,391 | 9,128 | -48% | 1 | 1 | 0% | 3,764 | 3,099 | -18% | 0 | 0 | — |
case-01 | fail→pass | 7,941 | 20,086 | +153% | 1 | 1 | 0% | 573 | 4,395 | +667% | 0 | 0 | — |
case-02 | fail→pass | 21,920 | 15,197 | -31% | 1 | 1 | 0% | 3,209 | 3,800 | +18% | 0 | 0 | — |
case-03 | fail→pass | 21,493 | 10,305 | -52% | 1 | 1 | 0% | 3,253 | 3,503 | +8% | 0 | 0 | — |
case-04 | fail→pass | 10,442 | 9,976 | -4% | 1 | 1 | 0% | 1,877 | 3,423 | +82% | 0 | 0 | — |
case-05 | fail→fail | 15,190 | 10,334 | -32% | 1 | 1 | 0% | 3,018 | 3,321 | +10% | 0 | 0 | — |
case-06 | fail→pass | 10,360 | 9,998 | -3% | 1 | 1 | 0% | 1,933 | 3,336 | +73% | 0 | 0 | — |
case-07 | fail→fail | 18,023 | 16,339 | -9% | 1 | 1 | 0% | 2,849 | 4,073 | +43% | 0 | 0 | — |
case-08 | pass→pass | 15,374 | 8,212 | -47% | 1 | 1 | 0% | 2,385 | 2,411 | +1% | 0 | 0 | — |
case-10 | fail→fail | 16,891 | 11,948 | -29% | 1 | 1 | 0% | 3,002 | 3,698 | +23% | 0 | 0 | — |
case-11 | pass→pass | 18,517 | 22,396 | +21% | 1 | 1 | 0% | 2,905 | 4,161 | +43% | 0 | 0 | — |
case-12 | pass→pass | 21,337 | 12,405 | -42% | 1 | 1 | 0% | 2,744 | 3,195 | +16% | 0 | 0 | — |
case-13 | fail→pass | 7,578 | 8,873 | +17% | 1 | 1 | 0% | 1,248 | 2,939 | +135% | 0 | 0 | — |
case-14 | fail→pass | 11,540 | 8,060 | -30% | 1 | 1 | 0% | 1,641 | 2,978 | +81% | 0 | 0 | — |
case-15 | pass→pass | 12,236 | 12,949 | +6% | 1 | 1 | 0% | 2,196 | 3,343 | +52% | 0 | 0 | — |
case-16 | fail→pass | 14,042 | 7,132 | -49% | 1 | 1 | 0% | 2,829 | 2,753 | -3% | 0 | 0 | — |
case-17 | fail→pass | 20,376 | 7,357 | -64% | 1 | 1 | 0% | 2,950 | 2,590 | -12% | 0 | 0 | — |
case-18 | fail→pass | 18,207 | 7,501 | -59% | 1 | 1 | 0% | 2,366 | 2,560 | +8% | 0 | 0 | — |
case-19 | fail→pass | 13,271 | 9,018 | -32% | 1 | 1 | 0% | 1,923 | 2,956 | +54% | 0 | 0 | — |
case-20 | fail→pass | 13,230 | 8,690 | -34% | 1 | 1 | 0% | 3,203 | 3,192 | -0% | 0 | 0 | — |
case-21 | fail→pass | 13,267 | 8,644 | -35% | 1 | 1 | 0% | 2,780 | 3,268 | +18% | 0 | 0 | — |
case-22 | fail→pass | 21,085 | 14,758 | -30% | 1 | 1 | 0% | 2,666 | 3,685 | +38% | 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 +64 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.