clonev

使用 Coqui XTTS v2 克隆任何语音并生成语音。超级简单 - 提供语音样本（6-30 秒 WAV）和文本，获取克隆语音音频。支持 14 种以上语言。当用户想要 (1) 克隆自己的声音或其他人的声音，(2) 生成听起来像特定人的语音，(3) 创建个性化语音消息，(4) 多语言语音克隆（使用克隆的语音讲任何语言）时使用。

# CloneV Skill - Voice Cloning Made Simple

## ⚠️ CRITICAL INSTRUCTIONS FOR AI MODELS

**DO NOT** try to use Docker containers directly.
**DO NOT** try to interact with `coqui-xtts` container - it is broken and restarting.
**DO NOT** try to use APIs or servers.

**ONLY USE THE SCRIPT:** `scripts/clonev.sh`

The script handles everything automatically. Just call it with text, voice sample, and language.

---

## What This Skill Does

Clones any voice from a short audio sample and generates new speech in that voice.

**Input**: 
- Text to speak
- Voice sample (WAV file, 6-30 seconds)
- Language code

**Output**: OGG voice file (cloned voice speaking the text)

**Works with**: Any voice! Yours, a celebrity, a character, etc.

---

## The ONE Command You Need

```bash
$(scripts/clonev.sh "Your text here" /path/to/voice_sample.wav language)
```

That's it! Nothing else needed.

---

## Step-by-Step Usage (FOR AI MODELS)

### Step 1: Get the required inputs
- Text to speak (from user)
- Path to voice sample WAV file (from user)
- Language code (from user or default to `en`)

### Step 2: Run the script
```bash
VOICE_FILE=$(scripts/clonev.sh "TEXT_HERE" "/path/to/sample.wav" LANGUAGE)
```

### Step 3: Use the output
The variable `$VOICE_FILE` now contains the path to the generated OGG file.

---

## Complete Working Examples

### Example 1: Clone voice and send to Telegram
```bash
# Generate cloned voice
VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "Hello, this is my cloned voice!" "/mnt/c/TEMP/Recording 25.wav" en)

# Send to Telegram (as voice message)
message action=send channel=telegram asVoice=true filePath="$VOICE"
```

### Example 2: Clone voice in Czech
```bash
# Generate Czech voice
VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "Ahoj, tohle je můj hlas" "/mnt/c/TEMP/Recording 25.wav" cs)

# Send
message action=send channel=telegram asVoice=true filePath="$VOICE"
```

### Example 3: Full workflow with check
```bash
#!/bin/bash

# Generate voice
VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "Task completed!" "/path/to/sample.wav" en)

# Verify file was created
if [ -f "$VOICE" ]; then
    echo "Success! Voice file: $VOICE"
    ls -lh "$VOICE"
else
    echo "Error: Voice file not created"
fi
```

---

## Common Language Codes

| Code | Language | Example Usage |
|------|----------|---------------|
| `en` | English | `scripts/clonev.sh "Hello" sample.wav en` |
| `cs` | Czech | `scripts/clonev.sh "Ahoj" sample.wav cs` |
| `de` | German | `scripts/clonev.sh "Hallo" sample.wav de` |
| `fr` | French | `scripts/clonev.sh "Bonjour" sample.wav fr` |
| `es` | Spanish | `scripts/clonev.sh "Hola" sample.wav es` |

Full list: en, cs, de, fr, es, it, pl, pt, tr, ru, nl, ar, zh, ja, hu, ko

---

## Voice Sample Requirements

- **Format**: WAV file
- **Length**: 6-30 seconds (optimal: 10-15 seconds)
- **Quality**: Clear audio, no background noise
- **Content**: Any speech (the actual words don't matter)

**Good samples**:
- ✅ Recording of someone speaking clearly
- ✅ No music or noise in background
- ✅ Consistent volume

**Bad samples**:
- ❌ Music or songs
- ❌ Heavy background noise
- ❌ Very short (< 6 seconds)
- ❌ Very long (> 30 seconds)

---

## ⚠️ Important Notes

### Model Download
- First use downloads ~1.87GB model (one-time)
- Model is stored at: `/mnt/c/TEMP/Docker-containers/coqui-tts/models-xtts/`
- Status: ✅ Already downloaded

### Processing Time
- Takes 20-40 seconds depending on text length
- This is normal - voice cloning is computationally intensive

---

## Troubleshooting

### "Command not found"
Make sure you're in the skill directory or use full path:
```bash
/home/bernie/clawd/skills/clonev/scripts/clonev.sh "text" sample.wav en
```

### "Voice sample not found"
- Check the path to the WAV file
- Use absolute paths (starting with `/`)
- Ensure file exists: `ls -la /path/to/sample.wav`

### "Model not found"
The model should auto-download. If not:
```bash
cd /mnt/c/TEMP/Docker-containers/coqui-tts
docker run --rm --entrypoint "" \
  -v $(pwd)/models-xtts:/root/.local/share/tts \
  ghcr.io/coqui-ai/tts:latest \
  python3 -c "from TTS.api import TTS; TTS('tts_models/multilingual/multi-dataset/xtts_v2')"
```

### Poor voice quality
- Use clearer voice sample
- Ensure no background noise
- Try different sample (some voices clone better)

---

## Quick Reference Card (FOR AI MODELS)

```
USER: "Clone my voice and say 'hello'"
→ Get: sample path, text="hello", language="en"
→ Run: VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "hello" "/path/to/sample.wav" en)
→ Result: $VOICE contains path to OGG file
→ Send: message action=send channel=telegram asVoice=true filePath="$VOICE"
```

```
USER: "Make me speak Czech"
→ Get: sample path, text="Ahoj", language="cs"  
→ Run: VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "Ahoj" "/path/to/sample.wav" cs)
→ Send: message action=send channel=telegram asVoice=true filePath="$VOICE"
```

---

## Output Location

Generated files are saved to:
```
/mnt/c/TEMP/Docker-containers/coqui-tts/output/clonev_output.ogg
```

The script returns this path, so you can use it directly.

---

## Summary

1. **ONLY use the script**: `scripts/clonev.sh`
2. **NEVER** try to use Docker containers directly
3. **NEVER** try to interact with the `coqui-xtts` container
4. Script handles everything automatically
5. Returns path to OGG file ready to send

**Simple. Just use the script.**

---

*Clone any voice. Speak any language. Just use the script.*

---

## 中文说明

# CloneV 技能 - 让语音克隆变得简单

## ⚠️ 给 AI 模型的关键说明

**不要**尝试直接使用 Docker 容器。
**不要**尝试与 `coqui-xtts` 容器交互——它已损坏并正在不断重启。
**不要**尝试使用 API 或服务器。

**只使用脚本：** `scripts/clonev.sh`

该脚本会自动处理一切。只需用文本、语音样本和语言来调用它。

---

## 此技能的功能

从一段简短的音频样本中克隆任何语音，并用该语音生成新的语音。

**输入**：
- 要朗读的文本
- 语音样本（WAV 文件，6-30 秒）
- 语言代码

**输出**：OGG 语音文件（以克隆语音朗读文本）

**适用于**：任何语音！你自己的、名人的、角色的等等。

---

## 你需要的唯一命令

```bash
$(scripts/clonev.sh "Your text here" /path/to/voice_sample.wav language)
```

就是这样！不需要其他任何东西。

---

## 分步用法（给 AI 模型）

### 第 1 步：获取所需输入
- 要朗读的文本（来自用户）
- 语音样本 WAV 文件的路径（来自用户）
- 语言代码（来自用户，或默认为 `en`）

### 第 2 步：运行脚本
```bash
VOICE_FILE=$(scripts/clonev.sh "TEXT_HERE" "/path/to/sample.wav" LANGUAGE)
```

### 第 3 步：使用输出
变量 `$VOICE_FILE` 现在包含生成的 OGG 文件的路径。

---

## 完整可运行示例

### 示例 1：克隆语音并发送到 Telegram
```bash
# Generate cloned voice
VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "Hello, this is my cloned voice!" "/mnt/c/TEMP/Recording 25.wav" en)

# Send to Telegram (as voice message)
message action=send channel=telegram asVoice=true filePath="$VOICE"
```

### 示例 2：用捷克语克隆语音
```bash
# Generate Czech voice
VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "Ahoj, tohle je můj hlas" "/mnt/c/TEMP/Recording 25.wav" cs)

# Send
message action=send channel=telegram asVoice=true filePath="$VOICE"
```

### 示例 3：带检查的完整工作流
```bash
#!/bin/bash

# Generate voice
VOICE=$(/home/bernie/clawd/skills/clonev/scripts/clonev.sh "Task completed!" "/path/to/sample.wav" en)

# Verify file was created
if [ -f "$VOICE" ]; then
    echo "Success! Voice file: $VOICE"
    ls -lh "$VOICE"
else
    echo "Error: Voice file not created"
fi
```

---

## 常用语言代码

| 代码 | 语言 | 示例用法 |
|------|----------|---------------|
| `en` | 英语 | `scripts/clonev.sh "Hello" sample.wav en` |
| `cs` | 捷克语 | `scripts/clonev.sh "Ahoj" sample.wav cs` |
| `de` | 德语 | `scripts/clonev.sh "Hallo" sample.wav de` |
| `fr` | 法语 | `scripts/clonev.sh "Bonjour" sample.wav fr` |
| `es` | 西班牙语 | `scripts/clonev.sh "Hola" sample.wav es` |

完整列表：en, cs, de, fr, es, it, pl, pt, tr, ru, nl, ar, zh, ja, hu, ko

---

## 语音样本要求

- **格式**：WAV 文件
- **时长**：6-30 秒（最佳：10-15 秒）
- **质量**：音频清晰，无背景噪声
- **内容**：任何语音（实际说的内容无所谓）

**好的样本**：
- ✅ 某人清晰说话的录音
- ✅ 背景没有音乐或噪声
- ✅ 音量一致

**差的样本**：
- ❌ 音乐或歌曲
- ❌ 强烈的背景噪声
- ❌ 太短（< 6 秒）
- ❌ 太长（> 30 秒）

---

## ⚠️ 重要注意事项

### 模型下载
- 首次使用会下载约 1.87GB 的模型（一次性）
- 模型存储在：`/mnt/c/TEMP/Docker-containers/coqui-tts/models-xtts/`
- 状态：✅ 已下载

### 处理时间
- 根据文本长度需要 20-40 秒
- 这是正常的——语音克隆是计算密集型的

---

##