Ai Voice Cloning

Preparing Your Audio Sample

Step 1 — Record or source audio

• Use a condenser microphone or a modern smartphone in a quiet room
• Avoid rooms with echo — record in a carpeted room or with sound panels
• Maintain consistent distance from mic (20–30 cm)

Step 2 — Clean the audio (if needed)

• Remove background music or ambient noise using a DAW (Audacity is free)
• Cut out long silences (keep natural pauses, remove dead air over 3 seconds)
• Normalize loudness to around -14 LUFS

Step 3 — Content variety Include sentences with:

• Questions and statements (intonation variety)
• Short and long sentences (pacing variety)
• Words with different vowel and consonant patterns
• Natural emotion (avoid robotic monotone reading)

Quality Checklist Before Uploading

• Single speaker throughout the entire file
• No background music, TV, or environmental noise
• No clipping (audio peaks going into red)
• At least 1 full minute of actual speech (not silence)
• Consistent recording environment (no room change mid-file)

Ethical Considerations

Voice cloning carries serious ethical responsibilities:

You must only clone:

• Your own voice
• A voice for which you have explicit written consent from the speaker
• Public figures only for clearly satirical or educational contexts where local law permits

Never clone a voice to:

• Impersonate someone without their knowledge
• Create fraudulent audio for scams, deepfakes, or misinformation
• Bypass consent in any commercial or broadcast context

ElevenLabs Terms of Service require you to confirm you have the rights to clone any voice you upload.

Using the Cloned Voice

After cloning, you receive a voice_id. Use it with:

/pocket-knife:elevenlabs-tts
  TEXT: "Your script here"
  VOICE: [your_voice_id]

The voice_id persists in your ElevenLabs account — you do not need to re-clone for each use.

Troubleshooting Poor Quality

Workflow Example

1. Record 3 minutes of natural speech as WAV
2. Run /pocket-knife:elevenlabs-voice-cloner with the file path and a name
3. Save the returned voice_id
4. Use /pocket-knife:elevenlabs-tts with that voice_id to generate narration
5. Reuse the same voice_id for consistent brand voice across projects