Stable Diffusion Prompt Guide

Nerdy Rodent
30 Aug 202211:33

TLDRIn this video from More Nerdy Rotten Geekery, the host explores the impact of specific words on image outputs using Stable Diffusion, a text-to-image generation model. By using consistent settings and the same seed, the video demonstrates how variations in prompts, such as 'focused,' 'sharp,' and 'painting,' can alter the resulting images. The experiments reveal how certain words can be more influential ('charcoal drawing') compared to others ('sharp'), and how word order and punctuation also affect the visuals. This insightful guide serves as a practical demonstration of 'prompt engineering' in the realm of AI-generated art.


  • ๐Ÿ”„ Using the same seed and text for a prompt results in deterministic output, meaning the image will be identical each time.
  • ๐Ÿ“ Adding or changing words in the prompt can significantly alter the generated image, even if the changes are subtle.
  • ๐Ÿ–Œ๏ธ The word 'focused' did not make the image more focused, showing that the impact of a word may not align with its literal meaning.
  • ๐Ÿ” The word 'sharp' may introduce minor changes, but it's not always clear whether it enhances sharpness as expected.
  • ๐ŸŽจ Using 'painting' as a prompt clearly changes the style of the image to resemble paintings.
  • ๐Ÿ“š The term 'chalk art' transforms the images into chalk art versions, maintaining the original structure.
  • ๐Ÿ’ก 'Concept art' has a medium impact, subtly changing the structure and style of the images.
  • ๐Ÿ“ท 'Canon m50', a camera model, strongly influences the image to look like a photograph, preserving the structure.
  • ๐Ÿ”Ž 'Close-up' works as expected, making the subjects appear larger and more zoomed in.
  • โœ๏ธ 'Charcoal drawing' is a very powerful word, drastically changing the structure to resemble charcoal drawings.
  • ๐Ÿ” 'Intricate' adds more detail to the images, making them more complex without significantly altering the structure.
  • ๐Ÿ”‘ The order of words in a prompt matters, with words closer to the beginning appearing to have a stronger impact.
  • โœ… Punctuation, such as commas and full stops, can also influence the output, sometimes adding backgrounds or changing details.
  • ๐Ÿ”ข The 'scale' parameter can affect the color saturation and clarity of the image, with higher values potentially leading to overblown colors and blurriness.

Q & A

  • What is the significance of using the same seed for running the same prompt twice in the context of stable diffusion?

    -Using the same seed ensures that the only variable changing is the prompt itself, allowing for a clear comparison of how different words or phrases affect the output image.

  • What effect did adding the word 'focused' have on the image generated by the stable diffusion model?

    -Adding the word 'focused' did not make the image more focused as expected, but it did introduce changes such as extra squiggles, a different hat shape, and different eyes.

  • Did the word 'sharp' make the images generated by the stable diffusion model noticeably sharper?

    -The word 'sharp' might have slightly changed the images, but the increase in sharpness was not significant enough to be easily noticeable.

  • How did the word 'painting' influence the output of the stable diffusion model?

    -The word 'painting' had a strong effect, making the generated images resemble paintings rather than photographs, with noticeable changes in style.

  • What was the impact of using the term 'chalk art' in the prompt?

    -The term 'chalk art' transformed the images into chalk art versions, maintaining the same structure but altering the style significantly.

  • Did the word 'concept art' significantly change the images generated by the model?

    -The word 'concept art' had a medium impact, with some images changing noticeably while others remained closer to the original photograph.

  • What happened when the camera model 'Canon M50' was used as a word in the prompt?

    -Using 'Canon M50' as a word resulted in all images being transformed into photographs, indicating a very strong influence on the output.

  • How effective was the word 'close-up' in changing the generated images?

    -The word 'close-up' was effective, as it resulted in closer, more zoomed-in images, although not necessarily sharper or more focused.

  • What was the impact of the word 'charcoal drawing' on the images generated by the stable diffusion model?

    -The word 'charcoal drawing' was very powerful, changing the structure and style of all images to resemble charcoal drawings.

  • How did the word 'intricate' affect the level of detail in the images?

    -The word 'intricate' added more detail to the images, making them more complex and intricate, although its power was not as strong as some other words.

  • How important is the order of words in a prompt when using a stable diffusion model?

    -The order of words in a prompt is important, as words closer to the beginning of the phrase seem to have more influence on the generated image.

  • What role does punctuation play in the generation of images using a stable diffusion model?

    -Punctuation can significantly affect the output, with changes such as adding a full stop or removing a comma resulting in different images.

  • How did adjusting the scale parameter in the prompt influence the colors and clarity of the generated images?

    -Increasing the scale parameter led to more overblown and blurry colors, while higher scales also caused significant changes in the images, including different poses and objects.



๐Ÿ–Œ๏ธ Prompt Engineering in Stable Diffusion: Word Impact

The first paragraph discusses the impact of specific words on image generation using stable diffusion. The speaker runs the same prompt twice with a cyberpunk cat wearing a steampunk hat, using the same seed for consistency. They then modify the prompt by adding words like 'focused', 'sharp', 'painting', 'chalk art', 'concept art', 'trending', 'canon m50', 'close-up', and 'charcoal drawing' to observe the changes in the generated images. The speaker notes that some words have a strong impact, like 'painting' and 'charcoal drawing', while others like 'sharp' and 'focused' do not produce the expected results. The paragraph emphasizes the importance of word choice and experimentation in achieving desired visual outcomes.


๐Ÿ“ Building Composite Prompts and Word Order Significance

The second paragraph explores the concept of composite prompts, where multiple descriptive words are stacked to create a more complex and detailed image. The speaker demonstrates how the order of words in the prompt affects the outcome, with words closer to the beginning appearing to have a stronger influence. They also test the impact of punctuation, such as commas and full stops, finding that even small changes in punctuation can lead to significant alterations in the generated images. The paragraph concludes with an encouragement for viewers to experiment with different words and share their findings in the comments.


๐Ÿ” The Effect of Scale on Image Generation

In the third paragraph, the focus shifts to the effect of scale on image generation. The speaker adjusts the scale from 10 to 30 and observes how it impacts the color saturation and clarity of the images. At lower scales, the colors are well-balanced, but as the scale increases, the colors become overblown, and the images start to appear blurry. The speaker suggests that playing with the scale can be a way to counteract overly saturated colors, and possibly combining it with text prompts to achieve a better balance. The paragraph ends with a call for viewers to share their experiences with prompt engineering and the words they find most effective.



๐Ÿ’กStable Diffusion

Stable Diffusion refers to a type of machine learning model that generates images based on textual input, or 'prompts.' In the video, the presenter experiments with different prompt variations to observe how slight changes can alter the generated images. This exploration helps viewers understand how specific words can influence the output of AI-driven image creation.


In the context of the video, 'prompts' are textual inputs given to an AI model to generate images. The video demonstrates how altering single words within prompts can significantly change the visual output, indicating the sensitivity and nuanced understanding of language by the AI.


Cyberpunk, as used in the video, is a genre of science fiction that focuses on futuristic technologically advanced societies, often featuring dystopian elements. The video utilizes this term in a prompt to create an image of a cat with a thematic aesthetic, demonstrating how genre-specific terms can guide the AI to generate images with certain stylistic elements.


Steampunk is a retrofuturistic subgenre of science fiction or science fantasy that incorporates technology and aesthetic designs inspired by 19th-century industrial steam-powered machinery. In the video, 'steampunk' is used to modify the hat of the cyberpunk cat, illustrating how adding specific thematic words to prompts can influence the AI to adjust details in its image generation.

๐Ÿ’กDeterministic output

The term 'deterministic output' in the video refers to the ability of the AI to produce the exact same image each time a prompt is run with the same settings and seed. This concept is highlighted to explain the reproducibility and consistency of results when using AI for image generation.


In the video, a 'seed' is a parameter used in the AI's process that ensures the reproducibility of images. By using the same seed with different prompts, the presenter demonstrates how changes in the input affect the output, while the same seed ensures that other variables remain constant.

๐Ÿ’กCharcoal drawing

The term 'charcoal drawing' is used in a prompt to transform the visual style of the generated images to resemble drawings made with charcoal. This demonstrates the AI's ability to interpret artistic mediums and apply their visual characteristics to generated images.

๐Ÿ’กConcept art

Concept art refers to the type of artwork used to visualize ideas before they are developed into final products, typically in video games, movies, or animations. In the video, the term is used to examine how the AI adjusts its rendering style, impacting the creativity and detail in the images.


The word 'intricate' in the video is used to assess whether adding this descriptor to a prompt enhances the complexity and detail of the generated images. This test helps to understand how descriptive adjectives influence the AI's interpretation and execution in image details.


In the video, 'scale' is discussed in the context of adjusting the resolution or detail level in the generated images. The presenter explores how different scale settings affect the clarity, color saturation, and overall composition of the images, providing insight into how this parameter can be manipulated to achieve desired visual effects.


Exploring the impact of specific words in stable diffusion prompts on the generated images.

Using 'focused' in a prompt doesn't necessarily make the image more focused, but changes the image's details.

The word 'sharp' slightly alters the image's sharpness, but the effect is minimal.

Including 'painting' in the prompt results in images that resemble paintings rather than photographs.

Using 'chalk art' in a prompt effectively transforms images to resemble chalk art.

Adding 'concept art' to a prompt makes slight modifications, but does not drastically change the style.

The phrase 'trending on Art Station' causes significant changes in the image, indicating its moderate strength.

Specifying a camera model like 'Canon M50' in the prompt converts images into photograph-like styles.

The word 'close-up' successfully makes images more zoomed in and detailed.

Including 'charcoal drawing' results in a strong transformation into charcoal art style.

The word 'intricate' adds more detail to the images, showing its effectiveness.

Composite prompts combining multiple keywords like 'charcoal drawing intricate concept art' create detailed and stylized images.

Word order in prompts influences the prominence of effects; words placed earlier have stronger influence.

Punctuation in prompts, such as full stops, significantly alters the visual outcome of images.

Adjusting the scale in prompts affects the color saturation and clarity, with higher scales causing images to appear overblown and blurry.