CFG SCALE

Last updated: June 20, 2025, 07:51 | Written by: Anthony Di Iorio

Imagine you're directing an artist.You give them a description of what you want them to paint. ですが、CFG Scale「1」よりは良い結果となっています。 CFG Scale「6」になると、クオリティも高くプロンプト内容に沿った画像生成もされやすくなっていました。 CFG Scale「10」でも「6」と同じくらいのクオリティを保っています。The CFG scale, or Classifier-Free Guidance scale, in Stable Diffusion is essentially how loudly you're giving those directions. put it as the last node on the model wire, and the higher you want to pump your CFG, the higher you'll set this. Seems that setting it to 0 is very similar to bypass, and setting it to 1.0 has the strongest affect, allowing you to crank CFG (though I'd still getting max 20, wheras on YouTube I saw people getting up to 50).It's a crucial setting that dictates how closely the AI image generator follows your text prompt. The Guidance Scale, also known as the Classifier-Free Guidance (CFG) scale, controls how closely Stable Diffusion adheres to the provided text prompt during the image generation process. In other words, it determines the extent to which the generated image reflects the input text. Impact of Guidance Scale on Image QualityToo loud, and the artist might become overly rigid, churning out something technically correct but lacking creativity.Too quiet, and they might wander off into a completely different direction, producing something beautiful but unrelated to your initial request.Understanding this delicate balance is key to unlocking the full potential of AI image generation.

This article delves deep into the world of the CFG scale, exploring its impact on image quality, how it interacts with other settings like sampler steps, and how to fine-tune it for different models and prompts. CFG guidance scale. This parameter can be seen as the Creativity vs. Prompt scale. Lower numbers give the AI more freedom to be creative, while higher numbers force it to stick more to the prompt. The default CFG used on OpenArt is 7, which gives the best balance between creativity and generating what you want.We'll explore the sweet spots, the potential pitfalls, and offer practical tips to help you master this essential tool. Scale (CFG) and Steps, the rest, leave as default these values seem dependent on one another As far as have seen, adding too many Steps to the default values either 'overcook' the process or do nothing; it seems that SD either gets confused, or thinks it got right by earlier steps.So, whether you're a seasoned AI art creator or just starting your journey, get ready to unlock the secrets of the CFG scale and elevate your creations to a whole new level.

What is the CFG Scale?

At its core, the CFG scale in Stable Diffusion is a parameter that controls the influence of your text prompt on the generated image.It's sometimes referred to as the Guidance Scale or Classifier-Free Guidance (CFG) scale, but they all refer to the same fundamental concept. Rendered using various steps and CFG values, Euler a for the sampler, no manual VAE override (default VAE), and no refiner model. All images were generated at . This is using the 1.0 version of SDXL. Summary: Subjectively, steps look best, with higher step counts generally adding more detail.It tells Stable Diffusion how much ""guidance"" to take from your written instructions when crafting an image.

Think of it as a knob you can turn to adjust the AI's ""obedience."" A higher CFG scale means the AI will try harder to match every detail in your prompt.A lower CFG scale allows the AI to be more creative and interpret the prompt more loosely.

How Does the CFG Scale Work?

The CFG scale works by influencing the diffusion process, the heart of Stable Diffusion's image generation. CFG scale is a parameter that controls how strict the AI should follow the prompt in image generation. Learn how to choose the best CFG scale value according to the complexity of the prompt words and see the effect of different CFG scale on the same prompt.During this process, the AI starts with random noise and gradually refines it into an image based on your prompt.

The higher the CFG scale, the stronger the influence of your prompt on each refinement step.This pushes the image closer and closer to what you described.Conversely, a lower CFG scale allows the random noise to have more influence, leading to a more diverse and unexpected outcome.

The Impact of CFG Scale on Image Quality

The CFG scale significantly affects the quality of your generated images. Learn how CFG scale influences image generation in stable diffusion, a popular AI model for image generation. Find out the optimal CFG scale value for different prompts and avoid common mistakes.Understanding its impact is crucial for achieving the desired results. 前述の「CFG scaleによる違いと推奨値」の通り、CFG scaleを上げていくと、イラストにノイズが入ったり、破綻したりすることがあります。この原因は、CFG scaleを上げた結果、Sampling stepsが不足してイラストの生成が不完全になっている可能性があります。Here’s a breakdown:

High CFG Scale: Sticking to the Script

A high CFG scale (e.g., 12-20 or higher) forces the AI to adhere strictly to your prompt. In Stable Diffusion, CFG stands for Classifier Free Guidance scale. CFG is the setting that controls how closely Stable Diffusion should follow your text prompt. It is applied in text-to-image (txt2img) and image-to-image (img2img) generations. The higher the CFG value, the more strictly it will follow your prompt, in theory.The resulting image will closely resemble your description, but this can come at a cost.

Pros: High prompt adherence, detailed and specific results.
Cons: Potential for over-saturation, unnatural appearance, artifacts, and reduced creativity. My go-to first-attempt is always 60/15 that is, 60 Steps, 15.0 CFG Scale. Tells me almost everything I need to know about going from there, usually CFG down, and maybe even Steps up. And of course, there will always be that one thing you do that seems to work best entirely outside what you thought was 'a sensible approach'.The AI may try too hard to include every detail, leading to a cluttered and less visually appealing image.
Use Cases: Ideal when you need precise control over the image's content, such as recreating a specific scene or object.

Medium CFG Scale: Finding the Balance

A medium CFG scale (e.g., 7-11) offers a balance between prompt adherence and creative freedom.This range is often considered the sweet spot for many prompts and models.

Pros: Good balance between accuracy and creativity, generally produces aesthetically pleasing results.
Cons: May require some experimentation to find the optimal value for specific prompts.
Use Cases: Suitable for general image generation, character design, and scenes where some artistic interpretation is desired.

Low CFG Scale: Embracing Creativity

A low CFG scale (e.g., 2-6 or lower) allows the AI to exercise more creativity and deviate from the prompt. Working on finding the best SDV settings. Currently I can't see a reason to go away from the default 2.5 configuration setting. Motion Bucket makes perfect sense and I'd like to isolate CFG_scale for now to determine the most consistent value. cfg_scale: number [ 0 . 10 ] Default: 2.5. How strongly the video sticks to the original image.This can lead to unexpected and often beautiful results.

Pros: Higher creativity, unique and artistic outcomes, better image quality in some cases.
Cons: May deviate significantly from the prompt, unpredictable results.
Use Cases: Ideal for abstract art, exploring different styles, and generating images with a more artistic flair. The default CFG scale value serves as a starting point, ensuring stable diffusion with good balance and low noise. Higher CFG Scale = More alignment with input, but potential distortion. Lower CFG Scale = More creativity, better quality, but potential deviation from input. Here is a concise guide for choosing the best CFG scale value:Experimenting to discover new aesthetics.

CFG Scale and Sampler Steps: A Synergistic Relationship

The CFG scale doesn't work in isolation.It's intertwined with other settings, particularly the number of sampler steps.Sampler steps determine how many times the AI refines the image during the diffusion process. CFGスケール（Classifier Free Guidance Scale）は、近年話題のStable Diffusionという画像生成モデルにおいて重要な概念です。このスケールは、生成される画像がどの程度入力されたプロンプトや画像に忠実になるかを決定するパラメータです。A higher number of steps generally leads to more detailed and refined images.

However, the optimal number of steps depends on the CFG scale.If you're using a high CFG scale, you might need more steps to fully resolve the image and avoid artifacts. there are plenty of prompts that produce interesting results at very low cfg, or very high cfg, or very low steps, or specific intermediate number of steps, and so forth. attempting to generalize as this chart does is a doomed mission because they're just is no generalization for a 500 dimension construct like stable diffusion that fits into aConversely, a low CFG scale might not require as many steps, as the AI has more freedom to fill in the details on its own.

As one commenter shared, their go-to starting point is ""60 Steps, 15.0 CFG Scale."" This highlights the importance of considering these two parameters together.

As a general rule:

High CFG Scale: Consider increasing the number of sampler steps.
Low CFG Scale: You can often get away with fewer sampler steps.

CFG Scale and Different Stable Diffusion Models

It's crucial to note that the optimal CFG scale value can vary depending on the specific Stable Diffusion model you're using.Different models have been trained on different datasets and may respond differently to varying levels of guidance.

For example, a model fine-tuned for photorealistic images might perform best with a lower CFG scale to allow for more natural-looking details.On the other hand, a model trained for stylized art might benefit from a higher CFG scale to ensure it captures the desired style accurately.

Therefore, experimentation is key.Don't be afraid to try different CFG scale values and observe how they affect the output of your chosen model. CFG (classifier-free guidance) tells Stable Diffusion how much guidance to use from your text prompt when generating an image. Most interfaces default the CFG scale to 7-8, which is a nice balance. You don t want the CFG scale to be too high, it will just overcomplicate the image as the AI attempts to render every single word as a detail.Look for resources like the model card or community forums to find recommended settings for specific models.

Finding the Optimal CFG Scale: A Practical Guide

So, how do you find the perfect CFG scale for your needs? CFG Scale: The Main Performance. After rehearsal, it s time for the show. The CFG Scale is how you mix the final performance: Mid CFG (7 8): Singer A takes the lead, but Singer B still adds a touch of improvisation. You ll get a fairly faithful rendition of scenery, outdoors, tree with a pink flower near the path yet thereHere's a step-by-step guide:

Start with the Default: Most interfaces default to a CFG scale of 7-8.This is a good starting point for general image generation.
Consider Your Prompt: Elaborate and detailed prompts often benefit from a higher CFG scale, while short and vague prompts might work better with a lower value.
Experiment and Iterate: Generate the same image with different CFG scale values and compare the results. It might sound like you want to keep the guidance scale at the highest value, but it will actually have negative effects on your image generation if you do. Having a high CFG scale setting creates images that start to look over-saturated and unrealistic. Recommended settings: Realistic images: Use a CFG Scale of 2-5Pay attention to the level of detail, accuracy, and overall aesthetic appeal.
Adjust Sampler Steps: If you're using a high CFG scale, experiment with increasing the number of sampler steps to improve image quality.
Consult Model Documentation: Check the documentation or community forums for your chosen model to find recommended CFG scale settings.

Common Mistakes to Avoid with the CFG Scale

While the CFG scale is a powerful tool, it's easy to make mistakes that can negatively impact your results.Here are some common pitfalls to avoid:

Using Too High of a CFG Scale All the Time: Just because a high CFG scale gives you precise control doesn't mean it's always the best choice. Learn how to use CFG scale and distilled CFG to control how closely Stable Diffusion follows your prompt and how much it improvises. See examples, explanations, and tips for different CFG settings and styles.It can lead to over-saturation, unnatural details, and a lack of creativity.
Ignoring Sampler Steps: The CFG scale and sampler steps work together.Don't neglect to adjust the number of steps when experimenting with different CFG scale values.
Not Experimenting: The optimal CFG scale varies depending on the prompt, model, and desired aesthetic.Don't be afraid to try different values and see what works best for you.
Overcomplicating Prompts with High CFG Scale: As one source mentions, avoid prompts that are too complex with a very high CFG scale, as the AI may attempt to render every word as a detail, leading to undesirable results.

Distilled CFG: A More Advanced Technique

For those seeking even finer control over the image generation process, Distilled CFG offers a more advanced technique.This method involves training a smaller, faster model to mimic the behavior of a larger, more complex model. Classifier-free guidance (CFG) is a fundamental tool in modern diffusion models for text-guided generation. Although effective, CFG requires high guidance scales, which has notable drawbacks: Mode collapse and saturation; Poor invertibility; Unnatural, curved PF-ODE trajectoryBy using a distilled model, you can often achieve similar results with lower CFG scale values, leading to improved image quality and faster generation times.

As one source notes, higher values (3-4) in Distilled CFG Scale can be useful for prompt adherence, especially when dealing with complex scenes.For example: ""A photo of a woman riding a mule on the surface of Mars wearing a cowboy hat and firing an Uzi into the air at a flying saucer.""

The Future of CFG and AI Image Generation

The field of AI image generation is constantly evolving, and the CFG scale is likely to remain a crucial parameter for controlling the creative process.As new models and techniques emerge, we can expect to see even more sophisticated ways to influence and refine the output of these powerful tools.

Researchers are also exploring alternative methods to CFG, such as dynamic thresholding, which aims to decouple image quality from the CFG scale. What does CFG Scale do? Question Share Sort by: Best. Open comment sort options. Best. Top. New.This could lead to even greater control over image generation without sacrificing visual appeal.

Frequently Asked Questions About CFG Scale

Here are some frequently asked questions that can further illuminate the topic of CFG Scale.

What is the default CFG Scale?

Most interfaces default the CFG scale to 7-8, which is generally considered a nice balance.

What happens if my CFG scale is too high?

A CFG Scale that is too high can overcomplicate the image as the AI attempts to render every single word as a detail. 3. Distilled CFG Scale. Distilled CFG Scale is very important. Higher values (3-4) can be useful for prompt adherence if you're trying to get a complex scene like: A photo of a woman riding a mule on the surface of Mars wearing a cowboy hat and firing an Uzi into the air at a flying saucer.It can also lead to over-saturated and unrealistic images.

What is CFG in Stable Diffusion?

In Stable Diffusion, CFG stands for Classifier Free Guidance scale, and it controls how closely Stable Diffusion should follow your text prompt.

Conclusion: Mastering the CFG Scale for AI Art

The CFG scale is an essential tool for anyone working with Stable Diffusion and other AI image generators.By understanding its impact on image quality, its relationship with other settings like sampler steps, and its variations across different models, you can unlock the full potential of these powerful tools and create stunning and unique works of art.

Remember, the key is experimentation.Don't be afraid to try different CFG scale values and see what works best for your specific prompts and creative goals. CFG scale is a setting that controls how closely Stable Diffusion follows your text prompt in text-to-image and image-to-image generations. Learn how CFG affects the quality of output images, how to balance it with sampler steps and methods, and how to play with it online.By mastering the CFG scale, you can take control of the AI image generation process and bring your artistic visions to life.So, go forth and create!