Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →真正的本地语音合成服务，使用 Edge-TTS 引擎，零依赖、零配置、完全离线可用，支持多语言和多种音色
.claude/skills/anbeime-qwen3-tts-local/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 96% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -3% | 0% |
直接提出需求，无需任何配置。
"朗读这段文本，用温柔女声"
"生成英语配音"
"用日语朗读这段话"
"用男声朗读解说词"服务自动执行：
| 音色名称 | 性别 | 风格 | 适用场景 | |---------|------|------|---------| | zh-CN-XiaoxiaoNeural | 女 | 年轻活泼 | 广告、短视频 | | zh-CN-YunyangNeural | 男 | 沉稳 | 纪录片、企业宣传 | | zh-CN-XiaohanNeural | 女 | 知性 | 新闻播报 | | zh-CN-YunjianNeural | 男 | 深沉 | 电影旁白 | | zh-CN-XiaomengNeural | 女 | 温柔 | 有声书、故事 | | zh-CN-YunxiNeural | 男 | 活泼 | 游戏解说 |
| 音色名称 | 性别 | 风格 | 适用场景 | |---------|------|------|---------| | en-US-JennyNeural | 女 | 美式 | 商务对话 | | en-US-GuyNeural | 男 | 美式 | 新闻播报 | | en-GB-SoniaNeural | 女 | 英式 | 正式场合 | | en-GB-RyanNeural | 男 | 英式 | 文学朗读 |
"朗读这段文本：欢迎使用本地语音合成服务""用温柔女声朗读这段文本：今天天气真好""用英语朗读这段话：Hello, how are you?"
"用日语朗读这段话：こんにちは""用激昂的语调朗读这段解说词：让我们勇往直前！"
"用温柔的语调朗读这段故事：很久很久以前..."bash# 安装 edge-tts pip install edge-tts # 验证安装 edge-tts --help
Edge-TTS 工作原理：
与 API 调用的区别：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | fail→pass | 14,198 | 4,627 | -67% | 1 | 1 | 0% | 2,384 | 2,008 | -16% | 0 | 0 | — |
case-03 | fail→pass | 8,136 | 6,769 | -17% | 1 | 1 | 0% | 1,273 | 2,106 | +65% | 0 | 0 | — |
case-01 | fail→pass | 10,002 | 4,319 | -57% | 1 | 1 | 0% | 1,593 | 1,893 | +19% | 0 | 0 | — |
case-02 | fail→pass | 6,537 | 4,468 | -32% | 1 | 1 | 0% | 1,017 | 1,990 | +96% | 0 | 0 | — |
case-04 | pass→pass | 3,362 | 3,140 | -7% | 1 | 1 | 0% | 615 | 1,729 | +181% | 0 | 0 | — |
case-06 | fail→pass | 13,359 | 5,745 | -57% | 1 | 1 | 0% | 2,136 | 2,079 | -3% | 0 | 0 | — |
case-07 | pass→pass | 15,142 | 7,819 | -48% | 1 | 1 | 0% | 2,366 | 2,600 | +10% | 0 | 0 | — |
case-08 | fail→pass | 16,603 | 7,053 | -58% | 1 | 1 | 0% | 2,596 | 2,285 | -12% | 0 | 0 | — |
case-09 | pass→pass | 11,227 | 5,330 | -53% | 1 | 1 | 0% | 1,814 | 2,019 | +11% | 0 | 0 | — |
case-10 | pass→pass | 8,711 | 3,205 | -63% | 1 | 1 | 0% | 1,644 | 1,693 | +3% | 0 | 0 | — |
case-11 | pass→pass | 10,903 | 4,873 | -55% | 1 | 1 | 0% | 2,084 | 1,995 | -4% | 0 | 0 | — |
case-12 | pass→pass | 10,911 | 3,961 | -64% | 1 | 1 | 0% | 1,972 | 1,749 | -11% | 0 | 0 | — |
case-13 | pass→pass | 9,516 | 3,124 | -67% | 1 | 1 | 0% | 1,703 | 1,713 | +1% | 0 | 0 | — |
case-14 | pass→pass | 9,096 | 3,243 | -64% | 1 | 1 | 0% | 1,594 | 1,638 | +3% | 0 | 0 | — |
case-15 | pass→pass | 13,688 | 5,423 | -60% | 1 | 1 | 0% | 2,205 | 2,057 | -7% | 0 | 0 | — |
case-16 | pass→pass | 7,556 | 3,989 | -47% | 1 | 1 | 0% | 1,465 | 1,774 | +21% | 0 | 0 | — |
case-17 | pass→pass | 8,815 | 3,503 | -60% | 1 | 1 | 0% | 1,761 | 1,817 | +3% | 0 | 0 | — |
case-18 | pass→pass | 7,537 | 3,588 | -52% | 1 | 1 | 0% | 1,497 | 1,795 | +20% | 0 | 0 | — |
case-19 | pass→pass | 7,979 | 2,766 | -65% | 1 | 1 | 0% | 1,400 | 1,629 | +16% | 0 | 0 | — |
case-20 | pass→pass | 13,924 | 6,270 | -55% | 1 | 1 | 0% | 2,296 | 2,332 | +2% | 0 | 0 | — |
case-21 | fail→pass | 15,960 | 10,012 | -37% | 1 | 1 | 0% | 3,243 | 2,844 | -12% | 0 | 0 | — |
case-22 | fail→pass | 14,210 | 8,270 | -42% | 1 | 1 | 0% | 2,725 | 2,716 | -0% | 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 +36 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.