What AI Image Generator Should YOU Be Using??

Matt Wolfe
19 Oct 202348:29

TLDRThis video provides an in-depth comparison of various AI image generators, assessing their performance across different criteria such as accuracy, creativity, realism, and the ability to handle specific use cases like illustrations, logos, textures, and text incorporation. The generators evaluated include Mid Journey, Dolly 3, Firefly Image 2, Stable Diffusion XL, Google's generative search experience, and Idiogram. The video grades each tool on prompt adherence, user interface usability, and pricing, concluding with a recommendation of Leonardo as the best overall value and Idiogram for uncensored and free use, while noting Mid Journey's strengths in creativity and realism despite its cost and usability drawbacks.

Takeaways

  • 😲 There is a multitude of AI image generators available, each with its own strengths and ideal use cases.
  • 🔍 The video compares several AI image generators based on various criteria such as accuracy, creativity, realism, and usability.
  • 🎨 Mid Journey is praised for its high-quality images but falls short in text generation and usability.
  • 🚀 Dolly 3, referred to as the 'Mid Journey killer', shows promise in accuracy but has limitations in certain areas like tiling and censorship.
  • 🔥 Firefly Image 2 is competitive in creating illustrations but struggles with more complex prompts and text generation.
  • 🌐 Stable Diffusion XL offers high customizability but may not always adhere closely to prompts.
  • 🔍 Google's generative search experience is user-friendly and free but may have limitations in generating certain types of content.
  • 📈 Idiogram initially excelled in generating text within images but faces competition from other platforms that have caught up.
  • 💰 Price and usability are significant factors, with some tools offering free tiers and others requiring subscription fees.
  • 🏆 Leonardo emerges as a strong contender due to its balance of performance across various criteria and a user-friendly interface.

Q & A

  • What is the main purpose of the video discussed in the transcript?

    -The main purpose of the video is to compare various AI image generators, evaluate their performance across different criteria such as accuracy, creativity, realism, and usability, and provide recommendations for specific use cases.

  • Which AI image generators were tested in the video?

    -The AI image generators tested in the video include Mid Journey, Dolly 3, Firefly Image 2, Stable Diffusion XL, Google's generative search experience, and Idiogram.

  • What criteria were used to evaluate the AI image generators?

    -The criteria used to evaluate the AI image generators were accuracy, creativity, realism, illustrations, logos and vectors, textures and backgrounds, usability of the user interfaces, and pricing.

  • How did the video test the accuracy of the AI image generators?

    -The video tested the accuracy of the AI image generators by providing specific prompts and grading the generators on how accurately they followed along with the prompts.

  • What was the general outcome of the accuracy test for Mid Journey?

    -In the accuracy test, Mid Journey received a score of 5 out of 10 without using the 'raw' style, and a score of 5.5 when using the 'raw' style, indicating that it did not adhere very closely to the prompts.

  • How did Dolly 3 perform in the accuracy test?

    -Dolly 3 performed exceptionally well in the accuracy test, scoring a 9 out of 10, as it was able to closely adhere to the given prompts.

  • What was the video's approach to testing the creativity of the AI image generators?

    -The video tested the creativity of the AI image generators by providing broad and minimal prompts and then making a subjective judgment on which tools produced more creative and less generic images.

  • Which AI image generator was found to be the most creative in the video?

    -Mid Journey was found to be the most creative AI image generator, followed closely by Stable Diffusion XL and Leonardo.

  • What was the video's method for evaluating the realism of the AI image generators?

    -The video evaluated the realism of the AI image generators by using a prompt to create an image of a couple holding hands in front of the Eiffel Tower and assessing the realism of the people and the location.

  • Which AI image generator was considered the most realistic in the video?

    -Mid Journey, specifically when using the 'raw' style, was considered the most realistic, followed by Firefly Image 2 and then the non-raw version of Mid Journey.

  • How did the video handle the evaluation of illustrations, logos, and vectors created by the AI image generators?

    -The video tested the AI image generators' ability to create illustrations, logos, and vectors by providing specific prompts and assessing the quality and style of the generated images.

  • What was the conclusion of the video regarding the best AI image generator for creating logos and vectors?

    -The video concluded that Google's generative search experience was the best for creating logos and vectors, followed by Mid Journey and Idiogram.

  • How did the video address the issue of censorship in the AI image generators?

    -The video tested the censorship by attempting to generate images with celebrity faces or IP and logos to see if the AI image generators would generate them or reject them due to content policy restrictions.

  • Which AI image generator was found to be the least censored in the video?

    -Idiogram and Stable Diffusion (Leonardo) were found to be the least censored AI image generators, willing to generate most prompts without restrictions.

  • What was the final verdict of the video on the best overall AI image generator?

    -The final verdict of the video was that Leonardo offered the best value overall, scoring well across most categories but particularly excelling in creativity and realism, and having a good balance of features and cost.

Outlines

00:00

🤖 AI Image Generators Comparison

The script discusses various AI image generators, including Mid Journey, Dolly 3, Firefly Image 2, Stable Diffusion XL, Google's generative search experience, and Idiogram. The video aims to determine the best tool for specific use cases by evaluating them on accuracy, creativity, realism, and other criteria such as user interface, pricing, and text integration within images. The comparison begins with an accuracy test using specific prompts to assess how closely each tool follows the given instructions.

05:02

🎨 Testing AI Generators for Creativity and Realism

The script continues by testing the AI generators' creativity with minimal prompts and evaluates their ability to produce unique and colorful images. It also assesses realism by using a prompt featuring a couple in front of the Eiffel Tower. The results are compared, and the generators are scored based on their performance. Mid Journey and Dolly 3 show strong creativity, while Mid Journey Raw and Firefly 2 also perform well in realism.

10:03

🏛 Evaluation of Illustration and Logo Creation

The video script moves on to test the AI tools' capabilities in creating illustrations and logos. It uses specific prompts to evaluate how well each tool can generate anime-style images and simple flat vector logos. The results show that Mid Journey, Leonardo, and Firefly 2 are particularly adept at creating illustrations, while Google surprisingly performs the best in logo creation.

15:05

🌆 Tilable Textures and Backgrounds Assessment

The script then focuses on the AI generators' ability to create tilable textures and backgrounds. It tests the tools using a 'colorful circuitry' prompt and checks if the images can tile seamlessly. Mid Journey and Stable Diffusion (via Leonardo) successfully create tilable images, while other tools like Dolly 3, Bing Image Creator, and Idiogram struggle with this task.

20:06

📜 Text Integration within Images Test

The ability to integrate text within images is tested next. The script uses prompts that include specific phrases to be displayed on signs held by penguins in the images. Dolly 3, Google, and Idiogram demonstrate the capability to include text accurately, while Mid Journey and Firefly 2 struggle with this feature.

25:06

🛡️ Censorship and Usability Analysis

The script discusses the censorship policies of the AI tools, noting their ability or inability to generate images with celebrity faces or intellectual property. It also evaluates the usability of each tool, considering the user interface and the ease of generating images. Leonardo stands out for its customizability, while Mid Journey lags due to its Discord-based interface.

30:07

💰 Pricing and Overall Ranking of AI Generators

The video concludes with a discussion on pricing and an overall ranking of the AI image generators based on their performance across various criteria. Leonardo emerges as the best value with a high score due to its lack of censorship and strong performance in most categories. Mid Journey and Idiogram tie for second place, with Mid Journey excelling in creativity and realism but having usability and cost drawbacks, and Idiogram being free and uncensored but less accurate and creative.

Mindmap

Keywords

💡AI image generators

AI image generators refer to artificial intelligence software that can create images based on textual descriptions or other input data. They are the central theme of the video, as the script discusses various AI image generators available, comparing their features and capabilities. Examples from the script include 'mid Journey', 'Dolly 3', 'Firefly image 2', 'stable diffusion XL', 'Google's image generator', and 'idiogram', all of which are evaluated for different use cases.

💡Mid Journey

Mid Journey is an AI image generator mentioned in the script as being highly regarded by many people. It is used to create images that are evaluated for accuracy, creativity, realism, and other criteria. The script specifically mentions testing Mid Journey with different prompts to assess its performance, such as generating a 'photo of a green bus floating in space' and a 'sitting artist with a bucket hat painting a canvas of a three-headed monster'.

💡Dolly 3

Dolly 3 is another AI image generator highlighted in the script, which some people are calling the 'mid Journey killer'. It is tested for its ability to adhere to prompts and create accurate and creative images. The script notes that Dolly 3 performed exceptionally well in the accuracy test, particularly when used within Bing's image Creator, and was able to generate images that closely matched the given prompts.

💡Stable Diffusion XL

Stable Diffusion XL, often abbreviated as 'sdxl' in the script, is an AI image generator that is discussed for its high level of customization. It is tested alongside other tools, and the script notes that it performed well in creating tilable textures and backgrounds. The platform 'Leonardo' is mentioned as a favorite for creating images with Stable Diffusion, indicating the use of additional creative effects in the pipeline.

💡Google's image generator

Google's image generator is integrated into their generative search experience, allowing users to generate images directly from search queries. The script mentions testing this tool and finding that it could generate images with certain prompts but struggled with others, such as creating an image representing 'beauty'. It is also noted for its free usage, which is a significant factor in the overall evaluation.

💡Idiogram

Idiogram is described in the script as a tool that was at the top of the AI art world about a month ago, known for generating text inside of images. It is evaluated for its ability to create images with text and is noted for its uncensored nature, as it does not seem to restrict the generation of copyrighted characters or celebrity likenesses.

💡Accuracy

Accuracy, in the context of the video, refers to how closely an AI image generator follows a given textual prompt to create an image. The script discusses testing the AI tools by providing specific prompts and then grading the images based on how accurately they represent the prompt. For example, the prompt adherence of generating a 'green bus floating in space' is used to evaluate this criterion.

💡Creativity

Creativity is assessed by providing minimal information in the prompt and judging the AI's ability to produce unique and interesting images. The script describes using vague prompts like 'beautiful, creative epic RGB image' to test how each tool interprets the request and generates images that are not just technically correct but also imaginative and artistic.

💡Realism

Realism is evaluated by testing the AI's ability to generate images that appear lifelike and could potentially be mistaken for photographs. The script uses the prompt 'image of a couple holding hands in front of the Eiffel Tower' to test this, noting the level of detail and authenticity in the generated images.

💡Illustrations

Illustrations are a specific type of image that are artistic and often used to explain or decorate text. The script discusses testing the AI tools' ability to create illustrations, particularly using the prompt 'anime girl with braids in the neon streets of Tokyo', and evaluates the style, contrast, and coherence of the images produced.

💡Logos and Vectors

Logos and vectors are tested to see how well the AI can create simple, flat, and potentially usable images for branding or design purposes. The script uses the prompt 'simple flat vector image logo of a wolf on a white background' to assess the AI's capability to produce clean, professional-looking designs.

💡Textures and Backgrounds

Textures and backgrounds refer to the AI's ability to create images that can be used as repeating patterns or seamless backgrounds. The script tests this by asking for 'colorful circuitry' and then checking if the images can tile without visible seams, which is crucial for certain design applications.

💡Text in Image

The ability to include accurate text within an image is a newer feature of AI image generators. The script tests this by using prompts that require text, such as 'a penguin holding a wooden sign that says subscribe to Matt wolf', and evaluates how well each tool can integrate the text into the image.

💡Censorship

Censorship in AI image generators refers to the restrictions on generating certain types of content, such as copyrighted material or images of specific people. The script tests this by using prompts that include celebrities and IP characters like 'Tom Hanks standing next to a stormtrooper' and notes which tools allow or block these images.

💡Usability

Usability pertains to how intuitive and easy-to-use an AI image generator is. The script evaluates the user interfaces of the different tools, noting the features and customizability each offers, such as aspect ratio changes, style modifications, and the ability to upload images for style matching.

💡Price

Price is an important factor when considering the use of AI image generators. The script discusses the cost of using each tool, noting whether they offer free tiers, the number of images included in paid plans, and the overall value for money, with some tools being free and others requiring a monthly subscription.

Highlights

Comparing multiple AI image generators to find the best tool for specific use cases.

Mid Journey, Dolly 3, Firefly Image 2, Stable Diffusion XL, Google's image generator, and Idiogram are evaluated.

Assessing generators on accuracy, creativity, realism, illustrations, logos, vectors, textures, backgrounds, usability, and pricing.

Mid Journey's raw style adheres more closely to prompts than the regular style.

Dolly 3 within chat GPT and Bing's image creator shows differences in image generation.

Stable Diffusion XL through Leonardo platform offers creative effects and customization.

Firefly Image 2's newest version shows promise but lacks in certain complex prompts.

Google's generative search experience has recently added image generation capabilities.

Idiogram's ability to generate text within images was a standout feature a month ago.

Accuracy testing shows Dolly 3's high score with a 9 out of 10.

Mid Journey excels in creativity, closely followed by Stable Diffusion XL and Leonardo.

Realism testing reveals Mid Journey Raw as the most realistic generator.

Illustration testing finds Mid Journey, Leonardo, and Firefly 2 as top performers.

Google's image generator stands out for creating logos and vectors.

Mid Journey and Stable Diffusion XL are capable of creating tilable textures and backgrounds.

Dolly 3, Google, and Idiogram are effective for incorporating text into images.

Censorship varies, with Idiogram and Stable Diffusion showing less restrictions.

Usability is highly rated for Leonardo, Firefly, and Dolly 3 within chat GPT.

Price considerations show a range of options from free to paid tiers.

Leonardo emerges as the best value with a balanced score across categories.

Dolly 3 within chat GPT underperforms due to cost and censorship issues.

The video concludes with recommendations based on different use cases for AI image generation.