ComfyUI Official
comfyanonymous
ComfyUI is a versatile graphical interface for building and managing complex Stable Diffusion workflows, utilizing a node-based system for enhanced user experience and flexibility.
87939
2025-09-09
VibeVoice Text to Speech Single Speaker
TTS
VibeVoice
1
51
Overview
VibeVoice generates expressive, context‑aware speech using an LLM plus a diffusion decoder, so it handles dialogue flow, turn‑taking, and emotion much more naturally than classic single‑speaker TTS. It uses continuous speech tokenizers at a very low frame rate (around 7.5 Hz), which lets it keep good audio quality while staying efficient enough to synthesize up to about 90 minutes of audio with as many as four different speakers in one pass.
Who can use it
VibeVoice is useful for:
Podcast and content creators who want AI‑generated multi‑speaker shows, interviews, or panel discussions from a script.
Audiobook and e‑learning producers needing long‑form narration with different character voices and natural prosody.
Developers building conversational agents, role‑play bots, or customer‑service simulations that require multiple distinct voices.
ComfyUI and pipeline users who want a free, high‑quality TTS node to add voiceover or dialogue on top of AI‑generated video.
Use case
A typical use case is writing a podcast script with labeled speakers (for example, Host, Guest1, Guest2), giving VibeVoice a short voice sample or style description for each, and generating a 30–60 minute audio episode with natural turn‑taking and expressive delivery. Another is feeding a course script into VibeVoice to create multi‑speaker e‑learning audio, where a main narrator explains topics and other voices ask questions or act out scenarios, ready to sync with slides or AI video.
Read more
N
EXTENSIONS
Nodes & Models
Overview
VibeVoice generates expressive, context‑aware speech using an LLM plus a diffusion decoder, so it handles dialogue flow, turn‑taking, and emotion much more naturally than classic single‑speaker TTS. It uses continuous speech tokenizers at a very low frame rate (around 7.5 Hz), which lets it keep good audio quality while staying efficient enough to synthesize up to about 90 minutes of audio with as many as four different speakers in one pass.
Who can use it
VibeVoice is useful for:
Podcast and content creators who want AI‑generated multi‑speaker shows, interviews, or panel discussions from a script.
Audiobook and e‑learning producers needing long‑form narration with different character voices and natural prosody.
Developers building conversational agents, role‑play bots, or customer‑service simulations that require multiple distinct voices.
ComfyUI and pipeline users who want a free, high‑quality TTS node to add voiceover or dialogue on top of AI‑generated video.
Use case
A typical use case is writing a podcast script with labeled speakers (for example, Host, Guest1, Guest2), giving VibeVoice a short voice sample or style description for each, and generating a 30–60 minute audio episode with natural turn‑taking and expressive delivery. Another is feeding a course script into VibeVoice to create multi‑speaker e‑learning audio, where a main narrator explains topics and other voices ask questions or act out scenarios, ready to sync with slides or AI video.
Read more