Using Craiyon and DeepAI to test text-to-art AI prompts | Why Try AI

Why Try AI?
22 Sept 202214:03

TLDRIn this episode of Why Try AI, the host introduces two free AI art generators, Craiyon and DeepAI, as tools for testing text-to-art prompts without registration. The video demonstrates how to refine prompts for better consistency in AI-generated images, despite the lower quality compared to paid models. The host guides viewers through the iterative process of prompt refinement, aiming for a dark, dystopian image of a giant mouse over a city, and highlights the importance of noting each attempt for future improvement.

Takeaways

  • 🎨 There are many AI art generators available that can create stunning artwork resembling the work of real artists.
  • 🔒 Many AI art generators require sign-up, registration, or even a waiting list, with some offering free credits to start.
  • 🆓 Introducing two free AI art tools: Craiyon (formerly Dali mini) and Deep AI, which require no registration for basic use.
  • ✍️ Both Craiyon and Deep AI focus users on crafting text prompts, which is a fundamental skill for interacting with AI art generators.
  • 🔄 These tools provide multiple images per text prompt, allowing users to test consistency and refine their prompts.
  • 🚫 The quality of images from Craiyon and Deep AI is lower than more sophisticated models, so they should not be used as a benchmark for AI capabilities.
  • 🔍 Consistency in results is more important than perfection when using these free tools to test and refine text prompts.
  • 📝 It's recommended to document each text prompt and the resulting images to analyze how changes affect the outcome.
  • 🤖 The process involves starting with minimal information and iteratively adjusting the text prompt based on the AI's output.
  • 🖼️ The example demonstrates creating a dark, dystopian image of a giant mouse over a city, using incremental adjustments to the text prompt.
  • 🚫 The video transcript warns that what works in Craiyon and Deep AI may not yield the same results in other AI models due to differences in algorithms.
  • ✨ The goal is to achieve a text prompt that consistently guides the AI to create the desired image, saving time and resources for more advanced tools.

Q & A

  • What is the main purpose of the video episode of 'Why Try AI'?

    -The main purpose of the video is to show viewers how to use free AI art generators to test out ideas, text prompts, and improve them for use in more sophisticated models.

  • What are some of the challenges with AI art generators that require sign up or payment?

    -Some AI art generators require sign up steps, a registration process, or even a waiting list. They may offer free credits initially, but eventually, users have to start paying to generate more images, which can be a barrier for casual users.

  • Which two free AI art generator tools are introduced in the video?

    -The video introduces Craiyon (formerly known as Dali mini) and Deep AI text to image generator as two free tools that allow users to generate art without signing up.

  • What is the core skill that the video suggests mastering when using AI art generators?

    -The core skill to master is crafting effective text prompts, as this is the primary way to communicate with AI art generators.

  • Why is it important to take note of the text prompts and images generated by AI art generators?

    -It's important to document text prompts and generated images to analyze how different tags added or removed affect the final outcome, which helps in refining the prompts for better results in the future.

  • What is the initial test case presented in the video?

    -The initial test case is to generate a dark dystopian movie poster of a giant mouse hovering over a city, with a sinister and possibly black and white aesthetic.

  • What is the first text prompt used in the test case and why was it insufficient?

    -The first text prompt was 'a giant mouse looking down on a city movie poster'. It was insufficient because it resulted in a cartoonish and laughable image, far from the desired sinister dystopian look.

  • What changes were made to the text prompt to achieve a more realistic and sinister look?

    -The text prompt was modified to include terms like 'photo of a giant mouse', 'from the ground', 'silhouette', and 'looming over a city' to achieve a more realistic and sinister look.

  • Why did adding 'laser eyes' to the text prompt not work as expected?

    -Adding 'laser eyes' did not work as expected because the AI interpreted it in a way that was not in line with the desired aesthetic, resulting in an image of a mouse having a bad trip at a rave instead of a futuristic sinister look.

  • How does the video suggest using the adjective 'intricate' in the text prompt?

    -The video suggests using 'intricate' to give the AI a cue to flesh out more details in the generated image, enhancing the quality and complexity of the artwork.

  • What was the final outcome of using the 'by Frank Miller' modifier in the text prompt?

    -The 'by Frank Miller' modifier did not improve the aesthetic as expected. It resulted in less consistency in the placement of the mouse in the images and did not add to the dark, black and white aesthetic.

  • What advice does the video give on refining text prompts for AI art generators?

    -The video advises to iterate on text prompts, removing elements that do not contribute positively to the desired outcome, and to not be afraid to discard ideas that do not work, as demonstrated by dropping the 'torso', 'laser eyes', and 'Frank Miller' modifiers.

Outlines

00:00

🎨 Introduction to AI Art Generators

The video begins with an introduction to the world of AI art generators, highlighting their ability to create stunning artwork that rivals human artists. The host encourages viewers to explore these tools despite the common requirement of sign-ups or payment for continued use. The focus shifts to two free AI art generators, Crayon (formerly known as Dali mini) and Deep AI Text to Image Generator, which offer a user-friendly experience without the need for registration. These tools allow users to input text prompts and receive multiple image outputs, helping to refine prompts for more sophisticated AI models. The host also warns about the lower quality of these free models compared to paid ones and emphasizes the importance of consistency in results rather than perfection.

05:02

📝 Iterative Process of Refining Text Prompts

The host demonstrates the iterative process of refining text prompts to achieve desired AI-generated images. Starting with a simple prompt about a giant mouse over a city, the host progressively adjusts the text to elicit a more realistic and sinister look. Key adjustments include removing the 'movie poster' aspect, adding terms like 'photo' and 'from the ground', and experimenting with phrases like 'silhouette' and 'torso'. The host also discusses the importance of noting down the prompts and the resulting images for future reference. Despite some missteps, such as adding 'laser eyes', the host gradually narrows down on a more effective prompt that captures the intended dystopian vibe.

10:04

🔍 Finalizing the Image Prompt and Testing Consistency

In the final part of the video, the host refines the text prompt further by adding descriptive adjectives like 'intricate' and 'scary', and experimenting with artist-specific styles like 'by Frank Miller'. The goal is to achieve a consistent and aesthetically pleasing result across multiple image outputs. The host finds that some additions, like 'laser eyes', do not improve the outcome and are subsequently removed. The process concludes with testing the final prompt on both Crayon and Deep AI to ensure consistency. The host is satisfied with the results and suggests that viewers can now take this refined prompt to more advanced AI models for better quality images, having saved time and potentially money in the process.

Mindmap

Keywords

💡AI art generators

AI art generators are software tools that use artificial intelligence to create visual art based on textual prompts. They are capable of producing images that can mimic the style of human artists, making them a powerful tool for artists and designers. In the video, the host discusses using these generators to test and refine text prompts, highlighting their potential for creating impressive artwork.

💡Craiyon

Craiyon, formerly known as Dali mini, is one of the free AI art generators mentioned in the video. It allows users to input text prompts and generates images based on those prompts without requiring any registration or sign-up process. The host uses Craiyon to demonstrate how to iterate and improve text prompts to achieve desired artistic results.

💡Deep AI

Deep AI is another free text-to-image AI generator highlighted in the video. Similar to Craiyon, it provides a platform for users to input text prompts and receive generated images. The video script mentions that Deep AI returns four images for each text prompt, which is useful for testing consistency and refining prompts.

💡Text prompt

A text prompt is a textual description provided to an AI art generator to guide the creation of an image. The effectiveness of the generated art heavily depends on the clarity and specificity of the text prompt. The video emphasizes the importance of mastering the art of crafting effective text prompts to communicate desired visual outcomes to AI art generators.

💡Image consistency

Image consistency refers to the uniformity or similarity in the output images generated by AI art generators in response to a specific text prompt. The video script discusses the importance of achieving consistency in the generated images to ensure that the text prompt effectively communicates the desired artistic vision.

💡Dystopian

Dystopian is a term used to describe a society characterized by oppression or control, often depicted in literature and art as a dark, grim, and oppressive environment. In the video, the host uses the term 'dystopian' in their text prompt to guide the AI in creating an image that conveys a sense of darkness and foreboding.

💡Sinister

Sinister is an adjective used to describe something that is threatening or evil in nature. In the context of the video, the host aims to create an image with a sinister feel, suggesting a dark and ominous atmosphere. This term is used to guide the AI in generating an image that evokes a sense of unease or danger.

💡Silhouette

A silhouette is the dark shape and outline of an object, space, or person visible against a lighter background. In the video, the host uses the term 'silhouette' in their text prompt to instruct the AI to create an image where the subject is depicted in this manner, adding to the dark and ominous aesthetic they are aiming for.

💡Photorealism

Photorealism is a style of art that aims to closely replicate the visual appearance of a photograph. The video script mentions the desire for a more photorealistic image, indicating a preference for images that look like they could have been taken with a camera, rather than being stylized or abstract.

💡Frank Miller

Frank Miller is a renowned comic book artist and writer, known for his dark, gritty, and highly detailed artwork, particularly in his work on 'Sin City'. In the video, the host attempts to use 'Frank Miller' as a modifier in their text prompt to guide the AI in creating an image with a similar aesthetic, suggesting a desire for a dark, black-and-white, and highly detailed visual style.

💡Laser eyes

Laser eyes is a term used in the video to describe a specific visual element the host initially wanted to include in their generated image. It refers to the depiction of eyes emitting beams of light, often associated with futuristic or science fiction themes. The video script shows the host experimenting with this term in their text prompt, but ultimately deciding it did not contribute positively to the desired outcome.

Highlights

Introduction to free AI art generators for testing text prompts.

AI art generators can create art that looks like it's drawn by a real artist.

Most AI art generators require sign-up or payment for more images.

Two free tools, Craiyon and DeepAI, allow testing without registration.

Craiyon and DeepAI generate multiple images per text prompt for consistency checks.

The quality of images from free tools is not as polished as paid models.

Consistency in image output is more important than perfection in free tools.

Different AI models may produce different results from the same text prompt.

Iterative testing with text prompts is crucial for mastering AI art generation.

Craiyon has a feature to screenshot all generated images with the prompt.

Starting with minimal information in the text prompt and iterating is recommended.

Example prompt: 'giant mouse looking down on a city movie poster'.

Adjusting the text prompt can significantly alter the generated image.

Adding terms like 'photo', 'from the ground', and 'silhouette' refines the image output.

Using adjectives like 'scary' and 'sinister' can enhance the desired aesthetic.

Inclusion of 'laser eyes' in the prompt led to an undesirable outcome.

Adding 'intricate' and 'by Frank Miller' to the prompt improved consistency.

Removing elements that do not contribute positively is a key strategy.

Testing the same prompt in both Craiyon and DeepAI can validate the results.

Iterative process with free tools saves time and potential costs in more sophisticated programs.