Multi Diffusion and ControlNet Tile Model Tutorial

Hello everyone! This is December!Let’s start today’s learning!Woohoo! Let’s take off!Today we are learning about the Multi Diffusion and ControlNet Tile modelFeels like I’m learning English…1. Using the Multi Diffusion Plugin

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion is an upgrade to the block processing of images in the upscale pluginCompared to Multi Diffusion, upscale has poorer controllability and slower drawing speedThe advantages of Multi Diffusion are good effects, high efficiency, strong controllability, and effectively reduces GPU consumptionLet me introduce the parameter settings

Multi Diffusion and ControlNet Tile Model Tutorial

In general, maintaining the default settings is fine to keep the width and height equal. If there is not enough video memory, reduce the size of the tile.

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion and ControlNet Tile Model Tutorial

Refer to the corresponding paper (I’ll research it when I have a chance)

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion and ControlNet Tile Model Tutorial

This is it

Note to lower the redraw amplitude a bit when using it, to around 0.3

Multi Diffusion and ControlNet Tile Model Tutorial

I also tried it

This is the original image

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion and ControlNet Tile Model Tutorial

After using

Multi Diffusion and ControlNet Tile Model Tutorial

It has indeed become much clearer

2. Generating 6k Images

If you need to generate a high-quality image, in addition to ultra-high resolution, you also need rich details

Multi Diffusion and ControlNet Tile Model Tutorial

This model can present a lot of details

Okay, let’s get started

First, use the TMND-mix model to generate an image

Multi Diffusion and ControlNet Tile Model Tutorial

You can see that the facial features of the person are very blurry

Second, fix the current seed, find the parameters below, copy and paste

Note not to add prompt words at the end, or the image will change

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion and ControlNet Tile Model Tutorial

Enable high-resolution restoration, choose the Latent series algorithm, which can express more details

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion and ControlNet Tile Model Tutorial

Set the redraw amplitude to around 0.5

Multi Diffusion and ControlNet Tile Model Tutorial

My computer can’t handle it a bit, the fan is whirring

Multi Diffusion and ControlNet Tile Model Tutorial

Third, then go to image-to-image and open multi diffusion

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion and ControlNet Tile Model Tutorial

The duration is shocking

My computer is struggling to run it, it crashed twice

Multi Diffusion and ControlNet Tile Model Tutorial

I’m exhausted, and produced such an image

Multi Diffusion and ControlNet Tile Model Tutorial

Adjusted this a bit higher

Multi Diffusion and ControlNet Tile Model Tutorial

Finally it ran out, I think it’s pretty good, my computer really struggles to run it

Seeing this code makes my heart clench Multi Diffusion and ControlNet Tile Model Tutorial

3. ControlNet Tile

In multi diffusion, chaotic images like this can easily occur

Multi Diffusion and ControlNet Tile Model Tutorial

Such scenes, while ControlNet Tile can avoid such problems

The principle is that ControlNet Tile has the ability to recognize content

Multi Diffusion and ControlNet Tile Model Tutorial

Multi Diffusion and ControlNet Tile Model Tutorial

The method of use is to open ControlNet Tile and Multi Diffusion simultaneously

However, it will increase the generation time by 3 to 4 times

Leave a Comment