the best REALISTIC models for Stable Diffusion

James Beltman
26 Jul 202308:44

TLDRThe video discusses the best models for generating hyper-realistic images using Stable Diffusion, highlighting 'Epic Realism' for its exceptional facial detail capture. Tips for optimizing results include using specific prompts, avoiding certain keywords, and utilizing upscaling techniques. 'Magic Mix' is recommended for dramatic scenes, while 'Analog Madness' excels at creating unique, non-stereotypical images based on vivid prompts. The video provides practical advice on samplers, steps, and conflict scale for achieving the best outcomes.

Takeaways

  • 🖼️ **Epic Realism**: This model excels at creating lifelike images with a focus on capturing facial details. It's the speaker's current favorite for its stunning results.
  • 🚫 **Prompt Simplicity**: Avoid using keywords like 'masterpiece', 'best quality', and '8K' as they don't affect the outcome. Instead, include terms like 'cartoon' and 'painting' in the negatives to maintain realism.
  • 🔍 **Fine-Tuning Parameters**: Keeping steps above 20 and adjusting CFG scale to 5 helps in balancing quality and realism. Different samplers like DPM sde Caris or dpm2m Keras can enhance realism.
  • 📈 **Resolution Improvement**: Using high res upscalers like nmkd super scale or nmkd faces with a denoising setting of 0.35 and an upscale factor of 2 improves image detail significantly.
  • 📉 **Negative Prompts**: Effective use of negatives can add realism and define what you don't want in the image, such as avoiding a bias towards East Asian women.
  • 💡 **Lighting Details**: The model captures intricate details like light shadows well, so there's no need for extra lighting keywords, and avoid 'cinematic' for a more natural effect.
  • 🎭 **Face Description**: Over-describing the face can lead to less desirable results, so it's better to keep facial descriptions minimal.
  • 🌟 **Epic Realism Helper**: Using the Epic realism helper by the same author can further enhance the creation of lifelike images.
  • 🔮 **Magic Mix Model**: Good for dramatic and dark scenes but has limitations in facial generation, often generating East Asian women with a slim face filter look.
  • ⚙️ **Sampler Options**: For Magic Mix, Eula a Euler, dpm2m Karis, or dpmsc cares work well, with an optimal number of steps between 20 and 40.
  • 🌐 **Analog Madness**: This versatile model can create images of ordinary individuals, with the power highly dependent on the vividness and robustness of the prompts provided.
  • 📝 **Crafting Prompts**: For Analog Madness, specific and pointed prompts work extremely well, and keywords like '3D Max', 'grotesque', and 'desaturated' can enhance realism.

Q & A

  • What is the title of the video transcript?

    -The title of the video transcript is 'the best REALISTIC models for Stable Diffusion'.

  • Which model is mentioned as the current favorite for creating lifelike images?

    -Epic Realism is mentioned as one of the current favorites for creating lifelike images.

  • What is the key to maintaining the perfect balance between quality and realism in the Epic Realism model?

    -The key to maintaining the perfect balance between quality and realism in the Epic Realism model lies in fine-tuning several parameters, including steps, CFG scale, and the chosen sampler.

  • What are the recommended settings for the high res upscaler to improve the level of detail on the generated image?

    -The recommended settings for the high res upscaler are a denoising strength of 0.35 and an upscale factor of 2.

  • How can one further enhance the realism of the images generated by the Epic Realism model?

    -One can further enhance the realism of the images by making effective use of negatives, avoiding biases towards creating East Asian women, and using the Epic Realism help addon by the same author.

  • What is the Magic Mix model known for in terms of image generation?

    -The Magic Mix model is known for its unique strengths in generating dramatic and dark lit scenes, bringing out moodiness and mystery in the images.

  • What are the limitations of the Magic Mix model when it comes to facial generation?

    -The limitations of the Magic Mix model in facial generation include a tendency to almost exclusively generate East Asian women and often leaning towards a uniform and unrealistic 'Tick-Tock slim face filter' look.

  • Which samplers are recommended for use with the Magic Mix model?

    -The recommended samplers for use with the Magic Mix model are Eula, Euler a, dpm2m Karis, or dpmsc cares.

  • What is the recommended range for the number of steps when using the Magic Mix model?

    -The recommended range for the number of steps when using the Magic Mix model is between 20 and 40.

  • What is the Analog Madness model known for in image generation?

    -The Analog Madness model is known for its versatility and dynamicism, with an ability to generate images of ordinary individuals, offering a refreshing alternative to the supermodel renditions.

  • What is the recommended workflow for using the Analog Madness model?

    -The recommended workflow for using the Analog Madness model includes using the sde Cara sampler, maintaining a range of 25 to 35 steps, and setting the conflict scale to the default of 7.

  • How can one optimize the prompts for the Analog Madness model to achieve better results?

    -To optimize the prompts for the Analog Madness model, one should provide vivid and robust prompts that are extremely specific and pointed, which work extremely well for this model.

Outlines

00:00

🎨 Epic Realism for Stunning Lifelike Images

The first paragraph introduces the Epic Realism model, which is praised for its ability to transform simple prompts into lifelike images with exceptional facial detail. The speaker shares tips for using the model effectively, such as keeping prompts simple and avoiding keywords that don't affect the outcome. They also discuss the importance of fine-tuning parameters like steps and CFG scale, and recommend specific samplers like DPM sde Caris or dpm2m Keras for enhanced realism. The use of high-resolution upscalers is also highlighted to improve image detail, with examples given to show the difference between upscaled and non-upscaled images. The paragraph concludes with advice on using negatives to guide the model away from biases and to define what is not desired in the generated image.

05:00

🌗 Magic Mix for Dramatic and Moody Scenes

The second paragraph focuses on the Magic Mix model, which is recognized for its strengths in creating dramatic and darkly lit scenes with a moody and mysterious atmosphere. However, it is noted that the model has limitations, particularly in generating diverse facial features, as it tends to produce East Asian women with a slim face filter look. The paragraph provides optimization tips for Magic Mix, including the use of different samplers and the ideal range for the number of steps and the CFG scale. The importance of upscaling for improved image quality is emphasized, with specific recommendations for upscalers and denoising settings. The paragraph also mentions the use of positive prompts and the impact of textual inversions on image quality, particularly in reducing cartoony appearances and improving hand depictions. The speaker concludes by acknowledging Magic Mix's unique style and its potential for crafting striking AI-generated artwork with atmospheric settings.

Mindmap

Keywords

💡Epic Realism

Epic Realism is a model for stable diffusion that is favored for its ability to transform simple prompts into lifelike images. It excels at capturing facial details that other models may overlook. In the context of the video, Epic Realism is used to demonstrate how fine-tuning parameters can maintain a balance between quality and realism, producing stunningly lifelike results.

💡Automatic 1111

Automatic 1111 refers to a specific version or setting within the stable diffusion software that the speaker uses to generate images. It is mentioned as a tool for creating firsthand examples of the images that can be made using the discussed models, emphasizing the practical application of the models in the video.

💡Prompts

Prompts are the inputs or instructions given to the stable diffusion model to generate specific types of images. The video emphasizes the importance of simplicity in prompts, suggesting that certain descriptive words like 'masterpiece' or '8K' do not affect the outcome, while others like 'cartoon' or 'painting' should be avoided to maintain realism.

💡CFG Scale

CFG Scale is a parameter within the stable diffusion settings that affects the level of detail in the generated images. The author recommends setting it to five to avoid compromising the realistic feel of the images. It is a crucial element in achieving the desired level of realism in the output.

💡Sampler

A Sampler in the context of stable diffusion models is an algorithm that determines how the model generates images from the prompts. Different samplers like DPM sde Caris, dpm2m Keras, and DPM fast are mentioned as options, each offering a unique approach to achieving realism in the generated images.

💡High Res Upscaler

High Res Upscaler is a tool used to improve the resolution of generated images, enhancing the level of detail. The video mentions nmkd super scale and nmkd faces as successful upscalers used with a denoising setting of 0.35 and an upscale factor of 2, which helps in achieving clearer and more detailed faces in the images.

💡Negatives

Negatives are terms or keywords that are included in the prompts to specify what should be avoided in the generated images. They are essential for adding realism and defining the undesired elements in the output. For instance, adding 'Asian, Chinese' as a negative helps to avoid a bias towards a specific ethnicity in the generated images.

💡Magic Mix

Magic Mix is another model for stable diffusion that is recognized for its strengths in creating dramatic and dark-lit scenes, enhancing the moodiness and mystery of the images. However, it has limitations, particularly with facial generation, often defaulting to East Asian women with a slim face filter look.

💡Analog Madness

Analog Madness is a versatile and dynamic model within stable diffusion that stands out for its ability to generate images of ordinary individuals, offering a refreshing alternative to the typical supermodel renditions. It emphasizes the importance of vivid and robust prompts to produce captivating outputs.

💡Steps and Conflict Scale

Steps and Conflict Scale are parameters within the stable diffusion settings that influence the detail and computational load of the generated images. The video suggests maintaining a range between 25 and 35 steps for optimal balance when using certain models like Analog Madness. The Conflict Scale, with a default setting of 7, is also highlighted as crucial for achieving the best results in terms of realism.

💡Nectar Prompts

Nectar Prompts are positive or negative prompts that guide the stable diffusion model to shape the desired and undesired aspects of the generated images. The video discusses the use of specific keywords like '3D Max', 'grotesque', and 'desaturated' to enhance realism in color and composition within the images.

Highlights

Epic Realism is a favorite model for creating lifelike images, especially capturing facial details.

Avoid adding extra keywords like 'Masterpiece' or '8K' to prompts as they don't affect the outcome.

Fine-tuning parameters such as steps and CFG scale is crucial for balancing quality and realism.

Using DPM sde Caris or dpm2m Keras samplers can enhance the realism of generated images.

High res upscalers like nmkd super scale or nmkd faces can significantly improve image detail.

Effective use of negative prompts can add realism and define what is not desired in the image.

Magic Mix model excels in creating dramatic and dark lit scenes with moody and mysterious atmospheres.

The Magic Mix model tends to generate East Asian women with a slim face filter look.

Optimal number of steps for Magic Mix is between 20 and 40 for best results.

Using terms like 'best quality', 'Masterpiece', and 'photorealistic' with Magic Mix can enhance image quality.

Analog Madness model is versatile and capable of generating images of ordinary individuals.

The potency of Analog Madness lies in the strength of the prompts provided.

SDE Cara sampler is the ideal choice for working with Analog Madness.

Keeping the conflict scale between 25 and 35 steps is recommended for Analog Madness.

Keywords like '3D Max', 'grotesque', and 'desaturated' can make images more realistic with Analog Madness.

Analog Madness can create a wide variety of realistic and unique images by playing with prompts.

Epic Realism, Magic Mix, and Analog Madness are models that offer different strengths in AI image generation.

Create Stunning AI Art Instantly

No login. No cost. Just incredible AI images !

Create Stunning AI Art InstantlyGenerate Now