Skip to main content
Voice Generator Node

​
Voice Generator

Give your words a voice.
This node turns text into natural-sounding speech. You can use a default voice, or clone your own (or someone else’s) by uploading audio samples. Perfect for narration, character dialogue, or any time you need spoken audio.

​
🪄 When to Use It

Use Voice Generator when you need spoken audio without recording. It works great for:
  • Voiceovers — narration for videos or presentations
  • Character dialogue — give your characters unique voices
  • Podcasts and audio content — scriptable speech
  • Dubbing — create audio in different voices
  • Prototyping — test dialogue before recording real actors
Skill level: Beginner
Average time: 10–30 seconds depending on text length
Cost: Low-Medium

​
🎚️ Controls and Parameters

​
Transcript (Text, Required)

The text you want spoken. Write it naturally, with punctuation for pacing. 💡 Tip: Add commas for pauses, periods for longer breaks. “Hello, how are you?” sounds more natural than “Hello how are you” Supports up to 2 speakers — label them in your transcript:

​
Speaker Sample Audios (Array, Required)

Upload 1-2 audio clips (5-15 seconds each) of the voice(s) you want to clone. 💡 Tip: Use clear, quiet audio with consistent tone. Avoid background music or noise.
  • 1 sample — clones one voice
  • 2 samples — supports dialogue between two different voices

​
Voice Speed (Slider)

How fast the speech should be:
  • 0.8 — slower, deliberate
  • 1.0 — normal conversational pace
  • 1.2 — faster, energetic
💡 Tip: Start at 1.0. Adjust if the pacing feels off for your use case.

​
CFG (Slider, 0–5)

Controls how closely the AI follows your voice sample:
  • Low (0.5–1.0) — more creative variation
  • Medium (1.3) — balanced (recommended)
  • High (2.0+) — very close match to sample
💡 Tip: Keep it around 1.3 unless the voice sounds too different from your sample.

​
Temperature (Slider, 0–1)

Controls speech variation and expressiveness:
  • Low (0.5–0.7) — consistent, monotone
  • Medium (0.95) — natural variation (recommended)
  • High (1.0) — more expressive, less predictable

​
Seed (Number)

Control randomness. Same seed + same inputs = same result.

​
🎨 Available Models

​
Vibe Voice (Default)

Advanced voice cloning and text-to-speech engine with natural intonation and emotion. Features:
  • Clone up to 2 different voices from audio samples
  • Multi-speaker support with speaker labeling
  • Fine control over speed, CFG, and temperature
  • Natural-sounding prosody and emphasis
💡 Tip: Vibe Voice handles all voice generation scenarios, from single-voice narration to multi-speaker dialogue.

​
🎨 What to Expect

Voice cloning captures the tone and timbre of your sample audio, but it’s not a perfect match.
The AI will:
  • Mimic the pitch and character of the voice
  • Follow the pacing and emphasis in your transcript
  • Handle most common words naturally
May struggle with:
  • Very long, complex sentences
  • Unusual names or technical terms
  • Extreme emotional range (shouting, whispering)
For best results:
  • Use clean audio samples with no background noise
  • Write clear, natural-sounding text
  • Keep sentences reasonably short

​
💬 Quick Tips

  • Record your voice samples in a quiet room with consistent tone
  • 10-15 seconds of sample audio is usually enough
  • Write your transcript the way you’d naturally speak it
  • Use punctuation to control pacing (commas = short pause, periods = longer pause)
  • For dialogue, clearly label speakers in your transcript
  • Generated audio matches the length of your text — longer text = longer processing time