Meigen Infinite Talk AI Generator

We have the newest lipsync model, Infinite Talk. It’s finally out, and I’m going to test it out. In this article, I’ll walk you through the entire workflow step by step and show you how to make some amazing content using it.

This model has a lot of potential for viral videos and creative projects. It’s definitely worth learning because it’s going to be a great tool to add to your toolkit.

Step 1: Download Models

As always, the model downloads are provided in the description section. We’ll quickly go through how to download one of them, and then you can use the same process for the rest.

For this walkthrough, I’ll only download the Infinite Talk model to save time. You can download the other models later in the same way.

Download Process

Click on the download link for Infinite Talk.
Once you click, the download should start automatically.
Some models are large, so depending on your internet speed, they might take a while.
Keep following the steps while your model downloads.

By the time you finish reading, your models should be ready for use.

Step 2: Placing the Model File

The description of the Infinite Talk model mentions where to place it. You need to put it inside the Comfy models diffusion models folder. Here’s how to do it:

Open your Comfy folder.
Navigate to:
```
models → diffusion models
```
Drag and drop the Infinite Talk model file into that folder.

That’s all you need to do. Follow the same process for the remaining models, and you’ll be ready for the workflow setup.

Step 3: Getting the Workflow

Now, it’s time to grab the workflows. In the resources, there’s a 1.2.2 text-to-image workflow available. We’ll use that to generate the first frame.

You can use any other workflow if you want — such as Flux or HiDream, depending on your preference. However, I recommend starting with just the Infinite Talk workflow first.

About the Workflows

Infinite Talk Fantasy Portrait: This workflow copies facial movements using lip sync. It requires a driving video.
The challenge is that royalty-free singing videos are not easy to find. So in this guide, I’ll show how it works, but we won’t have a perfect demo result.

The Infinite Talk workflow, however, will give us a much better generation result.

Step 4: (Optional) Artificial Studio Setup

You also have the option to use Artificial Studio, which is a custom-built tool I made. This step is completely optional.

If you want to use Artificial Studio:

Find the link in the description.
Once you open Artificial Studio, check the Infinite Talk box.
Then select lightex to voros.
Click Download.

That’s it. Again, this is not mandatory, but it can make the process smoother.

Step 5: Opening Infinite Talk Workflow in Comfy

Once everything is downloaded, open Comfy.

You can bring up the Infinite Talk workflow in two ways:

If you downloaded it manually, drag the file into Comfy.
If you’re using Artificial Studio, it should already appear in your available workflows.

Step 6: Preparing Audio and Image

Now we’re ready to generate content.

Audio Setup

Start by grabbing your audio file. You can use any royalty-free clip. For this example, I used one called “Falling for the Villain.”

Image Generation

Next, create a 2.2 image (text-to-image generation). For instance:

A blonde singer on stage singing into a microphone at a festival. She is wearing a white t-shirt and bell-bottom jeans.

Once your image is ready, drag and drop it into your Infinite Talk workflow.

Step 7: Ensuring Model Selection and Prompt Setup

Make sure you have all the required models selected — the ones you downloaded earlier.

Then add your prompt. Through experimentation, I noticed something interesting: using “arguing” instead of “singing” can sometimes produce more expressive results, adding passion to the facial expressions.

However, in this guide, I’ll stick to “singing” since we’re focusing on lip sync.

Step 8: Setting Time and Resolution

In the workflow, you’ll notice two groups: one for start time and one for end time.

Here’s how it works:

Enter start time in minutes and seconds.
Enter end time in minutes and seconds.
The system automatically calculates how long the clip is and how many frames it needs.

For example:

Start: 0:53
End: 1:03
This gives you a 10-second clip.

Resize Your Image

You can choose your preferred resolution:

720p
540p
480p

For this example, we’ll use 540p.

Once everything is set, you’re ready to run the workflow.

Step 9: Understanding What Happens Behind the Scenes

Let’s go over what each component in the workflow does:

Node Name	Function
Audio Crop Node	Cuts the audio to the selected length.
Audio Separation	Isolates vocals from the background music — great for those working in audio production.
Resize Image	Adjusts image size according to your desired resolution.
Clip Vision Encoding	Encodes the image similarly to the image-to-video process.
Infinite Talk Node	Processes image embeddings with context windows. Uses an 81-frame context window and overlaps 25 frames from the previous chunk.

The Infinite Talk node works in 81-frame chunks and overlaps 25 frames from the previous sequence to ensure smooth transitions. It generates at 25 frames per second (fps), so make sure your video combine setting matches that frame rate.

Step 10: Running Infinite Talk

Once you hit run, Infinite Talk begins processing your inputs — the image, audio, and parameters — and generates the synchronized video output.

When complete, the result is surprisingly realistic. On smaller screens like a phone, it would be difficult to tell it’s AI-generated. You can even create AI music videos, or simulate a virtual performer or pop star with it.

This shows how impressive Infinite Talk can be when used creatively.

Step 11: Infinite Talk + Fantasy Portrait Workflow

Next, let’s look at combining Infinite Talk with Fantasy Portrait.

I personally prefer using Infinite Talk on its own because it produces more consistent results. However, if you like experimenting, this combination can give some interesting outcomes.

How to Set It Up

Get your video ready (just like before).
Add your audio file — we’ll reuse the same one from earlier (“Falling for the Villain”).
Alternatively, use another clip. For this, I have one with a male singer.
Place your video and audio into the Fantasy Portrait workflow.

The goal here is to make the face movement follow the driving video while the mouth syncs to the audio.

Step 12: Important Note About This Method

Keep in mind:

This combination doesn’t always produce perfect results.
It works best when the audio and driving video match — for example, both showing and containing the same singer.

In this example, the driving video and audio are different. The system tries to adjust, but results may vary.

So, while you can experiment with this setup, I usually recommend sticking with Infinite Talk alone for best results.

Step 13: Testing the Combination

I ran the Infinite Talk + Fantasy Portrait setup.

Here’s what happens:

The face moves based on the driving video.
The mouth tries to sync with the given audio track.

While it does run successfully, I found the overall outcome less satisfying compared to Infinite Talk alone.

If you want consistent and high-quality lip-sync output, Infinite Talk alone is the preferred option.

Step 14: Final Output and Review

After testing both setups, here’s the summary:

Method	Description	Quality
Infinite Talk Only	Uses image and audio for direct lip-sync generation.	Excellent
Infinite Talk + Fantasy Portrait	Combines driving video with separate audio.	Moderate
Artificial Studio Setup	Optional helper tool for managing models and workflows.	Useful but optional

When played on a device, Infinite Talk’s generation looks smooth and natural. It’s ideal for creating AI-assisted music videos, voiceovers, or animated singing characters.

Step 15: Conclusion

That’s all for this walkthrough. We successfully covered how to:

Download and install the Infinite Talk model
Set up workflows in Comfy
Add audio and generate lip-synced outputs
Understand how each node functions
Experiment with Fantasy Portrait combinations

Infinite Talk is an exciting model for anyone looking to produce creative lipsync videos. With a bit of patience and experimentation, you can bring your static images to life with synchronized audio and expressive motion.