MultiTalk Lipsync in ComfyUI - New Update Support Multi-Characters Speaking Workflow

Tutorial Video : https://youtu.be/tDz8wEUoGnI In this vid

Tutorial Video : https://youtu.be/tDz8wEUoGnI

In this video, we dive into the latest update of the MultiTalk AI lipsync model and how it integrates with the WAN 2.1 AI video framework to create multi-character talking avatar videos. We are going to walkthrough a powerful workflow that uses Ollama/ Any LLM , Chatterbox Dialog TTS , and the updated Wan Video Wrapper to generate realistic conversations between multiple AI characters. Learn how to set up the tools locally or on cloud GPUs, use voice cloning for custom speaker identities, and apply segmentation masks to animate two or more characters independently. This in-depth tutorial also includes real-world testing results, performance benchmarks, and insights into current limitations—like audio sync issues and unnatural facial movements—while highlighting what works well and what still needs improvement.

Who is This Content Suitable For?

This content is ideal for:

- AI developers and creators working with talking avatars , lip-sync models , and AI-generated video content .

- Content creators interested in using ComfyUI , WAN 2.1 , and MultiTalk for podcast-style AI videos.

- Anyone experimenting with multi-speaker audio generation , voice cloning , and AI-driven character animation .

- Tech-savvy users looking to understand language models (like Ollama) , TTS systems , and video wrappers in practical workflows.

Why Does This Matter?

As AI continues to evolve, tools like MultiTalk and WAN 2.1 are pushing the boundaries of what’s possible in automated content creation , especially for podcasts , explainer videos , and interactive storytelling . This video gives you hands-on guidance on setting up and optimizing a complex but powerful pipeline, while also providing honest feedback about current limitations. Whether you're building AI agents, virtual hosts, or just exploring creative applications of generative AI, understanding how these models work together—and where they fall short—is crucial for making informed decisions in your own projects.

Resources and References :

• WanVideo Wrapper

  https://github.com/kijai/ComfyUI-WanVideoWrapper

• ComfyUI-Speaker-Isolation

  https://github.com/pmarmotte2/ComfyUI-Speaker-Isolation

• ComfyUI_Fill-ChatterBox

  https://github.com/pmarmotte2/ComfyUI_Fill-ChatterBox

• Multitalk model:

  https://huggingface.co/Kijai/WanVideo_comfy/blob/main/WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors

• Workflow: Ultimate Talking Avatar by Manu Le In MinicPC > Ultimate Talking Avatar

Update:

Attached workflow that I modify and use in local with Ollama version.