26 Jan 202306:01

TLDRIn this video, Binks introduces viewers to a photorealistic workflow using Stable Diffusion, a process they've been experimenting with recently. Binks shares their findings, emphasizing the impressive results they've achieved. They discuss transitioning from a keyword-based approach to a more structured English sentence format for prompts, inspired by their experience with large language models like GPT-3. Binks highlights the use of the DPM++ SD Kara sampler and the Realistic Vision version 1.2 model from Civet AI, noting its high-resolution capabilities and the need to be cautious of NSFW content on the site. They also mention the model's tendency to generate similar faces and drift from the original subject when given too much denoising strength. Binks provides tips for modifying prompts to achieve versatile results and shares how they've been using AI for world-building in a medieval fantasy game. The video encourages viewers to keep exploring AI and offers a helpful guide for those new to Stable Diffusion.


  • 🎨 The video discusses a photorealistic workflow using Stable Diffusion, a technique that the presenter has been experimenting with to achieve impressive results.
  • 📝 Binks shares that the video will be less of a tutorial and more about experimenting with Stable Diffusion settings, which will be provided in the comments.
  • 🔄 Binks has transitioned from a keyword-based approach to a more structured English sentence, inspired by large language models like GPT-3, which has yielded positive outcomes.
  • 🖼️ The DPM++ SD Kara sampler is the preferred choice for generating images, with a batch count of two and a resolution of 768x768 pixels.
  • ✨ The use of the Realistic Vision version 1.2 model from Civet AI is highlighted, which is known for producing high-quality images.
  • ⚠️ A caution is given about the presence of NSFW content on the Civet AI site, which can be disabled by users who prefer not to see it.
  • 🔗 Binks provides a link to a playlist of all his Stable Diffusion videos for further learning.
  • 🔍 The video demonstrates the versatility of the Realistic Vision model by modifying prompts and generating a variety of images.
  • 🧩 Binks mentions that the model sometimes generates similar faces, which could be due to the denoising strength or the freedom given to the model during image generation.
  • 🌐 The video emphasizes the potential of AI for world-building, particularly in the context of a medieval fantasy game that Binks is working on.
  • 💡 Binks encourages viewers not to get discouraged with Stable Diffusion, as it may take time to understand and master, but assures that he will continue to provide helpful content.

Q & A

  • What is the topic of the video?

    -The video is about creating photorealistic AI images using Stable Diffusion and a photorealistic workflow.

  • Who is the presenter of the video?

    -The presenter of the video is Binks.

  • What is the approach Binks has been experimenting with for Stable Diffusion prompts?

    -Binks has been experimenting with a more English structured sentence approach, similar to how large language models like GPT-3 work.

  • What sampler does Binks prefer to use with Stable Diffusion?

    -Binks prefers to use the DPM++ SD Kara sampler.

  • What is the resolution Binks sets for the images generated by Stable Diffusion?

    -Binks sets the resolution to 768 by 768 pixels.

  • Which version of Stable Diffusion does Binks recommend for photorealistic images?

    -Binks recommends using the Realistic Vision version 1.2 model from Civet AI.

  • What is the download size of the Realistic Vision version 1.2 model?

    -The download size of the Realistic Vision version 1.2 model is 3.8 gigabytes.

  • What is a potential issue Binks found with the Realistic Vision model?

    -Binks found that the model tends to generate similar faces, especially when using high denoising strength in image-to-image tasks.

  • How does Binks suggest modifying prompts for more versatile results?

    -Binks suggests changing the structure of the prompt to include different descriptions and characteristics to achieve more diverse outcomes.

  • What is Binks' advice for those who are new to using Stable Diffusion?

    -Binks advises not to get discouraged, as it takes time to understand and get used to Stable Diffusion, and to keep experimenting with it.

  • How does Binks use AI in his personal projects?

    -Binks uses AI for world-building, specifically for designing a medieval fantasy world for a game he is working on.

  • What does Binks encourage viewers to do if they have questions about the video content?

    -Binks encourages viewers to leave comments, like, and subscribe, and he will address their questions in future videos.



🎨 Experimenting with Stable Diffusion for Photorealistic Art

In this video, Binks introduces viewers to a new approach to using stable diffusion for creating photorealistic images. Binks shares his recent experiments with a more structured English sentence prompt, inspired by large language models like GPT-3, which has yielded impressive results. He demonstrates the process, including the use of the DPM++ SD Kara sampler, and provides settings details such as a batch count of two and a resolution of 768x768. Binks also mentions the need to download the specific model, Realistic Vision version 1.2 from Civet AI, and cautions about potential NSFW content on the site. He discusses the model's tendency to generate similar faces and the possibility of future updates addressing this issue. The video showcases the stunning results achievable with stable diffusion, emphasizing the model's versatility and potential for creative exploration.


🌐 Using AI for World Building and Creative Inspiration

Binks discusses how he has been utilizing AI for his hobby of world-building, specifically for a medieval fantasy game he is developing. He shares his enthusiasm for the creative potential of AI and encourages viewers to continue experimenting with stable diffusion, despite the learning curve. Binks offers to keep producing content on the topic and invites viewers to watch his other videos on stable diffusion, which have been found useful by many. He also encourages viewers to ask questions and engage with the content by leaving comments, liking, and subscribing to his channel.



💡Stable Diffusion

Stable Diffusion is a term referring to a type of artificial intelligence model used for generating images from textual descriptions. In the video, it is the central tool that the speaker, Binks, uses to create photorealistic AI images. The process involves experimenting with different settings and prompts to achieve the desired level of realism in the generated images.


Photorealistic refers to the quality of an image or visual representation that closely resembles a photograph. In the context of the video, Binks is aiming to produce AI-generated images that are so detailed and lifelike that they could be mistaken for actual photographs. This is a key goal of the workflow he is demonstrating.


A workflow in this context is a sequence of steps or processes that Binks follows to achieve the creation of photorealistic images using Stable Diffusion. The workflow includes setting up the AI model, crafting prompts, and adjusting parameters to refine the output.


In the context of AI image generation, a prompt is a text input that guides the AI in creating an image. Binks discusses how he has shifted from using simple keywords to more complex, structured sentences as prompts, which helps the AI generate more coherent and detailed images.

💡Negative Prompt

A negative prompt is a text input used in AI image generation to specify what should be avoided or not included in the generated image. Binks uses negative prompts to refine the output and prevent unwanted elements from appearing in the final images.

💡DPM++ SD Kara Sampler

The DPM++ SD Kara Sampler is a specific algorithm or method used within the Stable Diffusion model to generate images. Binks mentions it as his preferred choice for producing high-quality results.


Resolution refers to the clarity and detail of an image, often measured by the number of pixels in width and height. Binks sets the resolution to 768 by 768, which is slightly higher than the standard, to achieve a higher quality image output.

💡Denoising Strength

Denoising strength is a parameter in AI image generation that controls the level of noise or graininess in the final image. Binks discusses how setting this parameter too high can cause the AI to drift away from the original subject, which is something to be aware of when using Stable Diffusion.

💡Realistic Vision Version 1.2

Realistic Vision Version 1.2 is a specific model or version of Stable Diffusion that Binks is using in the video. It is noted for its ability to generate highly realistic images and is downloaded from Civet AI.

💡NSFW Content

NSFW stands for 'Not Safe For Work' and refers to content that may be inappropriate for professional settings. Binks warns viewers about the presence of such content on the site where the Realistic Vision model is downloaded and advises caution.


To upscale an image means to increase its size while maintaining or enhancing its quality. Binks suggests that the generated images from Stable Diffusion could be sent to an upscaler to further improve their resolution and detail.

💡World Building

World building is the process of creating an imaginary world, often for games, stories, or other creative projects. Binks mentions using AI for world building, specifically for designing a medieval fantasy world for a game, highlighting the utility of AI in generating inspiration and concepts.


Binks introduces a new photorealistic workflow with Stable Diffusion

The video will show settings and provide a prompt and negative prompt in the comments section

Experimenting with a large language model approach for prompts, inspired by GPT-3

Using DPM++ SD Kara sampler for image generation

Batch count set to two for higher resolution images

Image resolution set to 768 by 768 pixels

convex scale of seven is used for image detail

Restoring faces feature is checked for better facial details

Realistic Vision version 1.2 model from Civet AI is recommended for photorealism

Caution advised due to potential NSFW content on the Civet AI site

The model size is 3.8 gigabytes, suitable for the quality it provides

Tendency of the model to generate similar faces noted as a minor drawback

Future updates are expected to improve the model's performance

Stable Diffusion results are astonishingly good right off the bat

Modifying prompts to understand how Realistic Vision operates

Negative prompts effectively avoid unwanted image features

Demonstration of the model's versatility with different prompts

AI used for world-building and game design inspiration

Encouragement to keep experimenting with AI for fun and creativity

Binks will continue to provide content on Stable Diffusion