VibeVoice - Text to Speech

VibeVoice offers 6 speaker voices: Frank, Wayne, Carter, Emma, Grace, and Mike. Each voice has unique characteristics suitable for different content types, from narration to character voices.

CFG (Classifier-Free Guidance) scale controls how closely the generated speech adheres to the input text. Higher values (up to 3.0) increase text adherence, while lower values (as low as 0.5) allow more creative variation. The default value is 1.3 for balanced results.

The seed parameter allows you to control the randomization in generation. Using the same seed value with the same text will produce identical results, which is useful for reproducible generation and testing.

VibeVoice generates audio at 24kHz sample rate, providing high-quality, clear, and natural-sounding speech. The output is suitable for professional voiceover work and content creation.

VibeVoice is optimized for fast generation, making it suitable for real-time applications and interactive media. Generation speed depends on text length and server load, but typically completes in seconds.

Yes, you can use VibeVoice-generated audio for commercial projects including YouTube videos, podcasts, e-learning, audiobooks, advertisements, and more. Check the specific licensing terms for your use case.

VibeVoice supports long-form text input. For very long texts, consider splitting into multiple segments for optimal performance. The pricing is calculated per 1000 characters.

VibeVoice charges 6 credits per 1000 characters of text. Text under 1000 characters is rounded up to 1000. This makes it one of the most cost-effective TTS solutions available.