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 Multi Speaker
Multi Speaker
TTS
VibeVoice
1
38
VibeVoice for multi speaker is about using Microsoft’s VibeVoice models to generate long dialogues where several distinct voices talk, react, and take turns naturally in the same audio track.
Overview
VibeVoice is trained for multi‑speaker, long‑form text‑to‑speech, so it can handle multiple roles (host, guest, narrator, characters) in one pass, keeping each voice’s tone and rhythm consistent across the whole script. The model uses a language model plus a diffusion decoder and low‑rate speech tokens, which helps it capture context, pauses, and emphasis so conversations sound less robotic and more like a real recorded session.
Who can use it
Multi‑speaker VibeVoice is useful for:
Podcast creators generating full episodes with host and guests entirely from a written script.
Audiobook and drama producers who need different character voices and dialogue scenes without hiring several actors.
E‑learning and corporate training teams building scenario‑based conversations, role‑plays, and simulations.
ComfyUI and AI video users who want multiple characters speaking in sync with talking avatars or story videos.
Use case
A typical setup is to write a script with speaker tags (for example, “Host: …”, “Guest: …”), choose or define a voice style for each speaker, and let VibeVoice generate the entire multi‑speaker track in one go, preserving each voice across the episode. Another use case is an educational dialogue: two or three AI “teachers” and “students” explain a topic, ask questions, and respond to each other, producing a single audio file that can be synced to slides or AI‑generated classroom scenes.
Read more
N
EXTENSIONS
Nodes & Models
VibeVoice for multi speaker is about using Microsoft’s VibeVoice models to generate long dialogues where several distinct voices talk, react, and take turns naturally in the same audio track.
Overview
VibeVoice is trained for multi‑speaker, long‑form text‑to‑speech, so it can handle multiple roles (host, guest, narrator, characters) in one pass, keeping each voice’s tone and rhythm consistent across the whole script. The model uses a language model plus a diffusion decoder and low‑rate speech tokens, which helps it capture context, pauses, and emphasis so conversations sound less robotic and more like a real recorded session.
Who can use it
Multi‑speaker VibeVoice is useful for:
Podcast creators generating full episodes with host and guests entirely from a written script.
Audiobook and drama producers who need different character voices and dialogue scenes without hiring several actors.
E‑learning and corporate training teams building scenario‑based conversations, role‑plays, and simulations.
ComfyUI and AI video users who want multiple characters speaking in sync with talking avatars or story videos.
Use case
A typical setup is to write a script with speaker tags (for example, “Host: …”, “Guest: …”), choose or define a voice style for each speaker, and let VibeVoice generate the entire multi‑speaker track in one go, preserving each voice across the episode. Another use case is an educational dialogue: two or three AI “teachers” and “students” explain a topic, ask questions, and respond to each other, producing a single audio file that can be synced to slides or AI‑generated classroom scenes.
Read more