Local speech-to-text transcription skill based on OpenAI Whisper. Completely free, no API key required, supports Chinese and multiple languages. Ideal for private, offline voice transcription with high accuracy. Use when: 语音转文字, 语音转录, 本地STT, 语音识别, 音频转文字, speech to text, voice transcription, local STT, speech recognition, audio transcription, whisper transcription Related: voice-clone Part of UniqueClub toolkit. Learn more: https://uniqueclub.ai
Scanned 9/7/2026
Install to Claude Code
npx -y skills add wulaosiji/skills --skill whisper-stt --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Whisper Stt?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/wulaosiji-whisper-stt)More formats (shields.io, HTML) on the badges page.
---
name: whisper-stt
description: |
Local speech-to-text transcription skill based on OpenAI Whisper.
Completely free, no API key required, supports Chinese and multiple languages.
Ideal for private, offline voice transcription with high accuracy.
Use when: 语音转文字, 语音转录, 本地STT, 语音识别, 音频转文字,
speech to text, voice transcription, local STT, speech recognition,
audio transcription, whisper transcription
Related: voice-clone
Part of UniqueClub toolkit. Learn more: https://uniqueclub.ai
---
# Whisper STT 技能 (Whisper Speech-to-Text)
本地语音转文字,隐私安全,完全免费,无需 API Key。
## When to Use
**适用于以下场景:**
- 需要将语音/音频转换为文字
- 隐私要求高的场景(本地处理,不上传云端)
- 无需网络连接的场景
- 中文语音转录
- 多语言语音转录
- 与语音克隆结合使用(语音→文字→AI处理→语音)
**Do NOT use this skill if:**
- 需要实时流式转录(Whisper 适合离线批量处理)
- 对转录速度要求极高且设备性能有限
- 需要云端协作或共享转录结果
- 音频质量极差(背景噪音过大)
**触发关键词 / Trigger Phrases:**
- 语音转文字 / speech to text
- 语音转录 / voice transcription
- 本地STT / local STT
- 语音识别 / speech recognition
- 音频转文字 / audio transcription
- 转录音频 / transcribe audio
- whisper转录 / whisper transcription
## Workflow
### 1. 安装依赖
```bash
pip3 install openai-whisper
```
### 2. 使用封装脚本
```python
from tools.whisper_stt import transcribe
# 转录音频
result = transcribe("audio.mp3", model="base", language="zh")
if result.get("success"):
print(f"转录内容: {result['text']}")
else:
print(f"错误: {result['error']}")
```
### 3. 命令行使用
```bash
# 基本用法
whisper audio.mp3
# 指定中文和模型
whisper audio.mp3 --model base --language zh
# 指定输出格式
whisper audio.mp3 --model base --language zh --output_format txt
```
### 4. 与语音克隆结合
```python
from tools.whisper_stt import transcribe
from tools.voice_clone_api import generate_speech
# 语音转文字
stt_result = transcribe("user_voice.mp3")
text = stt_result["text"]
# AI处理...
response = f"收到: {text}"
# 文字转语音(克隆声音)
generate_speech(response, voice_id="wuna-001")
```
## Model Selection
| 模型 | 大小 | 速度 | 准确率 | 推荐场景 |
|------|------|------|--------|---------|
| tiny | 39MB | 最快 | 一般 | 测试 |
| base | 74MB | 快 | 良好 | **日常使用** ✅ |
| small | 244MB | 中等 | 好 | 平衡 |
| medium | 769MB | 较慢 | 很好 | 高质量 |
| large | 1550MB | 最慢 | 最佳 | 精确转录 |
## Guardrails
### 音频质量要求
- **格式**: MP3, WAV, M4A 等常见格式
- **清晰度**: 语音清晰,避免背景噪音
- **音量**: 正常音量,避免过小或失真
- **语言**: 支持中文、英文及 90+ 语言
### 限制说明
- **实时性**: 不适合实时流式转录
- **资源占用**: 大型模型需要较多内存
- **准确性**: 口音、方言可能影响准确率
- **时长**: 长音频可能需要分批处理
### 故障排除
| 问题 | 解决方案 |
|------|---------|
| 安装失败 | 先装PyTorch CPU: `pip3 install torch --index-url https://download.pytorch.org/whl/cpu` |
| 速度慢 | 换用tiny/base模型 |
| 中文不准 | 确认指定 `--language zh` |
| 内存不足 | 使用base/tiny模型 |
## Related Skills
| 技能 | 关系 | 说明 |
|------|------|------|
| [voice-clone](./voice-clone) | 配套 | 语音克隆技能,可与STT形成完整语音工作流 |
## About UniqueClub
Part of UniqueClub toolkit - AI-powered creative tools for speech processing.
Learn more: https://uniqueclub.ai
---
*基于OpenAI Whisper开源项目*

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!