AI image creation

Prompts

A specific instruction or input provided to an AI model to guide its generation of content, such as text or images. It serves as a starting point or guideline for the AI to understand what type of content to produce.


Tips for writting effective prompts:


  • Be Clear and SpecificUse descriptive language and provide detailed instructions to guide the AI's output. Avoid ambiguity or vague wording that could lead to misinterpretation.Instead of "Create an illustration of a forest", a prompt like "Create an illustration of a tranquil forest scene with sunlight filtering through the canopy and a winding stream flowing through the moss-covered rocks”, will provide specific details about the setting, mood, and elements to include in the illustration, giving the AI a clear direction for visualizing the scene. 
  • Provide ContextProvide context or background information to help the AI understand the desired outcome. Example: "Create an image for a sustainable fashion brand that incorporates eco-friendly symbols and imagery, reflecting the brand's commitment to ethical practices." It provides context by specifying the purpose and values of the brand. 
  • Use ExamplesInclude examples or references to illustrate the desired style, tone, or content. Example: "Generate an abstract painting inspired by Vincent van Gogh's 'Starry Night,' exploring themes of imagination, wonder, and the night sky." By referencing a famous artwork, this prompt inspires the AI to create a visually captivating painting in a similar style. 
  • Consider Constraints Consider any constraints or limitations that may apply to the task. For example, if generating images, specify the dimensions, aspect ratio, or color palette. If generating text, specify the desired length, format, or language style. In some of the platforms you can even upload or choose an image to match the style.
  • Use Chat GPT to improve your prompts.



Gemini

Examples


  • Can create illustrations but its difficult to adjust style:

  • Gemini is really good for creating images and adding texts into them like "mockups":

  • Gemini does not create images with people at the moment (4/29/2024): "We are working to improve Gemini’s ability to generate images of people. We expect this feature to return soon and will notify you in release updates when it does."

  • Gemini can't add company brand assets: "I understand that you'd like an image that includes the company brand but I can't generate images that directly reference real-world companies due to copyright restrictions."
  • Images are always 1:1
  • Editing options not available, you can give feebback to continue improving results but no further tools are available. Images are regenerated when given a feedback, it cannot directly modify an image already created.


Subscription information

-
 

Copyright information

Users retain full copyright ownership of images generated using Gemini in paid plans.


Resources

Upcoming updates:  Article link 
How to write better prompts in Gemini:  Doc Link 


Adobe Firefly



Examples


  • Illustrations are good for presenting a concept or idea, but are not that detailed:

  • Not good at creating images with text:

  • Can also create patterns:

  • Firefly has a  Generative Fill  tool. You can upload an image and edit it manually by selecting the part you want to change and typing a new prompt:

  • Once you are done you can download the photo or add text to it in Adobe Express:


  • Provides various text effects, fonts, and customization options to ensure that the added text seamlessly integrates with the image. 
  • Additional editing tools to adjust: ratio, style, color and tone, camera angle, lighting.
  • You can upload a reference image to match its outline and depth.


Subscription information

PIC already has a subscription.


Copyright information

Users retain copyright ownership of images edited or created using Firefly.


Resources

"Adobe Firefly vs. DALL·E " - Blog Post link 
"Adobe Firefly vs. Midjourney: How Firefly can help speed up creative workflows." -  Blog post link 
Updates in photoshop:  Blog post link 



DALL-E (OpenArt)



Examples



  • Illustrations are more detailed, icons don't break like they do in Firefly:

  • Not good at creating images with text:

  • It has a Sketch to Image tool that can create realistic images from a quick drawing:

  • Includes image editing and manipulation tools, allowing users to refine generated results to better fit their creative vision. 
  • Advanced features for controlling specific attributes of generated images, such as shape, texture, and lighting, through intuitive sliders and controls. 
  • You can upload a reference image to match its outline and depth.
  • Expand image: upload an image and change its size, ratio, quality. It will generate and fill.
  • Model Training: upload a few images to teach AI what you want and you will get your custom model. Only for paid subscription.


Subscription information

One-time 50 trial credits to experience all features. Then they offer 3 different subscriptions:



Copyright information

Users retain copyright ownership of images generated using DALL-E. You may use, modify, and distribute these images for commercial or non-commercial purposes, provided they are not used to create illegal or unethical content.


Resources

Inspiration of images created in Dall-e:  https://openart.ai/discovery 
"The Most Complete Guide to Stable Diffusion Parameters" -  Blog post link 



Midjourney



Examples

  • Illustrations are detailed, icons don't break.

  • Images with text not always work, error in the text appear sometimes:

  • Really good for creating patterns:

  • Midjourney is a little more complex than the other platforms. Here is a  Quick Start Guide  to understand how to begin creating images. 
  • It works with commands. For writting a prompt you should start with the command: /imagine prompt
  • All editing tools have specific commands to add to the prompt. For example, to specify the image ratio you should use the command --aspect <value>:<value>, or --ar <value>:<value>The prompt should look something like this: /imagine prompt vibrant California poppies --ar 5:4
  • Also can create short video of your images like this one: https://cdn.midjourney.com/b54871f6-0b76-424d-9cfe-128425b049b8/video.mp4 
  • Check the  Quick Start Guide  to see all the commands to edit your image.



Subscription information

PIC already has a subscription.


Copyright information

Free to use the images anywhere we want with that subscription.


Resources



Comparisons


PROMPT: Create a futuristic cityscape with towering skyscrapers, neon lights, and flying cars buzzing through the streets.

PROMPT: A flamingo made of watermelon

PROMPT: A surreal cityscape where buildings are made of candy and cotton candy clouds float in the sky
  • Gemini: combines elements of realism and fantasy. The resulting image might strike a balance between imagination and believability.
  • Adobe Firefly: highly polished and realistic. The resulting image would have a professional and cinematic quality, resembling a scene from a sci-fi movie.
  • DALL-E: Surreal and imaginative twist. The resulting image might have a dreamlike quality, with vibrant colors and surreal elements that push the boundaries of reality.
  • Midjourney: leans towards artistic and creative outputs. Has a high level of control and flexibility when it comes to generating images.



Conclusion


Title
Title
Title
Title
Title
Feature
Gemini
Adobe Firefly
DALL-E
Midjourney
Access
Text-to-Image
Yes
Yes
Yes
Yes
General style
Limited. Combines elements of realism and fantasy.
More options, including artistic and realistic. Highly polished and realistic images.
Extensive options, including artistic styles. Surreal and imaginative twist.
Extensive options, including artistic styles
Additional Editing Tools
No
More control over details and variations
Extensive control with advanced parameters
More control with advanced parameters and keywords
Edit ratio
No. Images have 1:1 ratio
Yes
Yes
Yes
Upload reference image
No
Yes
Yes
Yes
Add brand assets
No
No
No
No
Illustrations
Creates illustrations but its diffcult to adjust style.
Good for presenting a concept or idea, but are not detailed.
Good and more detailed
Good and more detailed
Text application into image
8/10
4/10
4/10
7/10
Images with people
No, at least for now.
Yes
Yes
Yes
Output Format
JPG, PNG
JPG, PNG
JPG, PNG
JPG, PNG
Extra tools
No
  • Text effects, fonts, and customization options
  • Generative Fill tool.
  • Integrates with other Adobe Creative Cloud apps
  • Creates a gallery with all your generated images.
  • Sketch to image
  • Expand Image
  • Model training
  • --tile parameter generates patterns
  • --video parameter generates short simple videos
  • Creates a gallery with all your generated images.

Subscription
PIC already has a subscription
Once the 50 credits trial ends, 10/month the basic plan
PIC already has a subscription
Commercial use
Yes
Yes
Yes
Strengths
Fast generation, simple interface, free tier
More realistic and detailed outputs
Highly creative and artistic outputs
High level of control, strong artistic styles
Disadvantages
  • Only 1:1 ratio
  • No extra editing tools
  • No images with people
  • Illustrations and icons are not detailed.
  • Text in image doesn't work
  • Text in image doesn't work
  • More complex to use, need to learn parameters in order to edit


At the moment is not possible to use brand assets in the AI image creation with any of these platforms.

It is recommended that the PIC team focus their efforts on:

    Firefly for its simple interface, easy to use editing tools and its generative fill tool. We already have an Adobe suscription so its included in that plan.
    Gemini for prompt creation and text in image purposes.

    Midjourney for more advanced creations and its high level of control. It requires more time to understand how to use it and learn parameters, but it has very detailed and advanced options. Ilustrations and patterns are really good.
    Dall-e for its sketch to image tool which the other platforms don't offer. PIC doesn't have a suscription and It's not necessary to have one, but with the 50 free credits it provides we can take advantage of this tool.


Appendix



AI integrations - Photoshop


  • Text-to-Image Generation: Create images based on descriptions you provide in text prompts.
  • Generative fill: Remove or replace objects within existing photos using AI
  • Generative expand: change the size of your artboard and it will generate the remaining parts.

  • Background Removal/Generation: Isolates your subject from the background with greater accuracy. Generates backgrounds for your images based on AI understanding of the scene.



AI integrations - Illustrator


  • Text to Vector: creates vector graphics from scratch based on text descriptions.

  • Recolor tool:


  • Retype Tool: can select outlined text or a an image with text in it, and it would find the font matches:

  • Mockup Tool: helps visualize your design in a real-world context wih adobe stock images: