How to Add LCM Sampling Method in Stable Diffusion

1. Introduction LCM LoRA's

In this blog, we explore integrating the LCM Sampling Method to elevate LCM LoRA, guaranteeing fast and top-notch image and video creation within stable diffusion. No need for an XYZ plot to identify the optimal sampling method; we'll guide you through two effective approaches to implement the LCM Sampling Method. Curious about LCM LoRA's? Check out the tutorial video below for a comprehensive introduction and better understanding.

2. Adding LCM Sampling Method with AnimateDiff Extension

Our initial method for incorporating the LCM Sampling Method into Stable Diffusion involves installing the 'AnimateDiff' extension. This extension guarantees the appearance of the 'LCM' sampling method in both the text2image and img2img tabs. Follow the steps below to seamlessly implement this extension:

Go to the "Extensions" tab located within Stable Diffusion.
Select the "Available" subtab.
Click the "Load from" button.
Look for "AnimateDiff" and proceed to click on the "Install" option

create-video-animations-with-animatediff-lcm-loras-a1111-faster-rendering-speed.webp

After completing the installation of the extension, head to the 'Installed' tab, and choose 'Apply and restart UI.' For a smooth experience and to address potential errors, I strongly suggest performing a full restart of your Stable Diffusion.

FAST Video Generation with AnimateDiff & LCM LoRA's (A1111)

The example below demonstrated the utilization of the AnimateDiff Extension in conjunction with LCM LoRA and the LCM Sampling Method.

3. Adding LCM Sampling Method with Code Snippets

Now, let's explore the second method for incorporating the LCM Sampling Method in Stable Diffusion. While the first approach involved installing the AnimateDiff Extension to introduce the LCM Sampling Method to the UI, it's important to note that the AnimateDiff Extension may potentially clash with other extensions.

For a more seamless option, we recommend adding a code snippet to implement the LCM Sampling Method directly. This way, you won't need to download the AnimateDiff Extension solely to access LCM Sampling Methods. Although AnimateDiff is an excellent extension to master, let's delve into the second method of adding the LCM Sampling method.

Adding the Code Snippet

You need to make two small edits using a text editor. Follow these steps:

Edit the file "sampling.py" located at the following path: "\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\sampling.py"
Add the following code snippet at the end of the file and save it:


@torch.no_grad()
def sample_lcm(model, x, sigmas, extra_args=None, callback=None, disable=None, noise_sampler=None):
    extra_args = {} if extra_args is None else extra_args
    noise_sampler = default_noise_sampler(x) if noise_sampler is None else noise_sampler
    s_in = x.new_ones([x.shape[0]])
    for i in trange(len(sigmas) - 1, disable=disable):
        denoised = model(x, sigmas[i] * s_in, **extra_args)
        if callback is not None:
            callback({'x': x, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i], 'denoised': denoised})

        x = denoised
        if sigmas[i + 1] > 0:
            x += sigmas[i + 1] * noise_sampler(sigmas[i], sigmas[i + 1])
    return x

The second change is done in the file "sd_samplers_kdiffusion.py" found here: "\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py"
Open the file and look for a list called "samplers_k_diffusion"
Add the following to the end of the list (before the "]") and save the file:


('LCM Method', 'sample_lcm', ['lcm'], {})

So in the end the "samplers_k_diffusion" list should look something like this:


samplers_k_diffusion = [
All your other sampling methods,
('LCM Method', 'sample_lcm', ['lcm'], {})]

After making those changes and saving the files, restart your Stable Diffusion. Once the UI reloads, you'll find 'LCM Method' in the list of Sampling Methods. Great! Now, choose 'LCM Method' when you want to speed up creating images/videos with LCM LoRA. This ensures top-notch quality.

4. Conclusion LCM Sampling Method

In conclusion, we've explored two effective approaches to integrate the LCM Sampling Method into Stable Diffusion for enhanced image and video creation. Whether through the straightforward installation of the 'AnimateDiff' extension or the manual addition of some code snippets, both methods empower users to access the LCM Sampling Method seamlessly. By following the provided steps, users can now choose the 'LCM Method' among the Sampling Methods options, revolutionizing their stable diffusion experience.