BETTER than PROMPTS - The Future of AI Composition
TLDRIn this video, the presenter, known as the king of AI, demonstrates advanced techniques for creating detailed and controlled AI-generated compositions. The focus is on creating a compelling image of a female hero with angel wings against a burning city backdrop. The process involves separating the subject from the background, manipulating light and color through curves adjustment, and integrating additional elements like wings and a cityscape. The presenter also shares tips on refining the image with Affinity Photo and further enhancing it using various settings in the AI tool, such as Euler, denoise strength, and CFG scale. The video concludes with a demonstration of another method combining a portrait with an abstract background, showcasing the versatility and creativity possible with AI composition tools.
Takeaways
- 🎨 **Image Composition Control**: The speaker emphasizes the need for more control over image composition elements, suggesting that traditional text prompts may not suffice for detailed creative work.
- 🖼️ **Describing with Images**: To better describe an image, the process involves using another image, highlighting the power of visual references in the creative process.
- 🚫 **Layer Separation**: The script demonstrates how to separate the foreground from the background in an image, a crucial step for detailed editing and composition.
- 🖌️ **Selection and Masking**: The use of selection tools and brushes is fundamental for refining the image and isolating specific elements, such as the woman and the rooftop.
- 🧚♀️ **Adding Elements**: The process of adding angel wings to a character is shown, illustrating how to integrate new visual elements into an existing composition.
- 🌆 **Background Manipulation**: The speaker discusses how to incorporate a burning city into the background, showing the importance of context in creating a compelling scene.
- 🔆 **Adjusting Brightness and Contrast**: The use of curves adjustment to modify the brightness and contrast of elements, such as the wings, is explained to achieve the desired visual effect.
- 📐 **Resolution and Export**: The importance of setting the correct resolution and exporting the final composition is highlighted, with specific settings provided for optimal results.
- 📝 **Prompt Refinement**: The script details how to refine prompts for AI to generate more accurate and detailed images, such as specifying a 'female hero with blonde hair'.
- 🔍 **Upscaling and Denoising**: Techniques for upscaling images and reducing noise are discussed, including the use of specific software features and settings.
- 🧩 **Combining Images**: The method of combining a portrait with an abstract background is demonstrated, showcasing how to blend different images to create a unique composition.
Q & A
What is the main theme of the composition the speaker wants to create?
-The speaker wants to create a composition featuring a female hero with angel wings standing on the top of a roof, with a burning city in the background.
How does the speaker suggest separating the foreground from the background in an image?
-The speaker suggests using the selection brush in add mode to select the woman and the rooftop, then refining the selection and applying it to separate the foreground from the background.
What is the process for adding angel wings to the female hero in the composition?
-The process involves opening an image with wings, copying the desired wing, pasting it into the main image, positioning it on the hero's back, duplicating the layer, flipping it horizontally, and adjusting the brightness and contrast of the wings using the Curves adjustment.
How does the speaker describe the method of adding a burning city to the background?
-The speaker describes the method as placing the burning city image in the composition, resizing it as needed, and positioning it behind the model and wings to create a specific scene with a nice contrast.
What are the settings the speaker uses for exporting the composition?
-The settings used for exporting include reducing the resolution on the long side to about 1600, using Euler with 25 steps, a restore phase of 512 by 768, a CFG scale of 7, a denoise strength of 0.4, and a batch size of eight.
How does the speaker propose to enhance the hero's face in the composition?
-The speaker proposes to send the image to inpaint to paint out the face of the character and then use a new prompt to generate a beautiful face, using a mask and denoising with a batch size of 8.
What is the method for combining a portrait with an abstract background?
-The method involves opening the portrait in Affinity Photo, placing the abstract background, and using the selection brush to remove the original background of the portrait. Then, the abstract background is re-enabled and positioned to complement the portrait.
How does the speaker use text to image for creating a composition?
-The speaker uses text to image by loading the portrait with the abstract background into Automatic 1111, using a prompt that describes the desired composition, and applying various control nets and settings to achieve the final result.
What is the sampling method used by the speaker for creating the composition?
-The sampling method used is DPM++ 2sa Keras with sampling steps set to 25.
How does the speaker adjust the brightness and contrast of the wings in the composition?
-The speaker adjusts the brightness and contrast by selecting both wings, grouping them, and then using the Curves adjustment to push the upper right part (bright areas) up and pull the lower part (dark areas) down to create an S curve.
What is the speaker's approach to fixing imperfections in the composition?
-The speaker's approach to fixing imperfections involves using the inpaint tool to address areas like the eyes and then upscaling the image for a refined final result.
Why did the speaker delete the part of the video where a bear holding a red balloon and a girl sitting in his lap were featured?
-The speaker deleted that part of the video by accident and offers to recreate it during a live stream if the viewers are interested.
Outlines
🎨 'Creative Image Composition with AI'
The first paragraph introduces a method for creating compelling compositions using AI, with a focus on having more control over elements such as composition, colors, and details. The speaker, referring to themselves as the 'king of AI,' guides the audience through a step-by-step process of creating an image of a female hero with angel wings against a burning city backdrop. The process involves separating the foreground from the background, adding angel wings to the character, adjusting the brightness and contrast, and placing a burning city in the background. The paragraph concludes with exporting the image and adjusting settings for resolution and composition details.
🖼️ 'Image Upscaling and Text-to-Image Techniques'
The second paragraph demonstrates advanced techniques for image manipulation and text-to-image conversion using AI. It begins with a discussion on combining a portrait with an abstract background in Affinity Photo, removing the original background, and integrating a new one. The process includes refining selections, masking, and exporting the image with specific dimensions. The paragraph then shifts to text-to-image conversion, showcasing how to input prompts and use control networks for depth and resolution, resulting in a composition of a beautiful African queen with a colorful abstract background. The speaker also mentions an accidental deletion of a part of the video and offers to recreate it in a live stream, ending with a call to action for likes and future engagement.
Mindmap
Keywords
💡AI Composition
💡Rasterize
💡Selection Brush
💡Refine
💡Wings
💡Burning City
💡CFG Scale
💡Denoising
💡Batch Size
💡Image to Image
💡Control Net
Highlights
The speaker introduces a new approach to AI composition that offers more control over elements, composition, and colors than traditional prompts.
Describing an image with another image is suggested as the best way to achieve detailed control in AI composition.
The process of separating the foreground from the background using selection and rasterization techniques is demonstrated.
Adding angel wings to a female hero figure is shown, with adjustments made for lighting and contrast.
The integration of a burning city background into the composition to create a dramatic scene is explained.
Exporting the composition with specific resolution and prompt parameters for further refinement is discussed.
The use of Euler, steps, restore phase, CFG scale, denoise strength, and batch size in settings for image generation is highlighted.
A method for painting out and improving the face of a character in the composition using a new prompt is shown.
Upscaling the image with tile resampling and control net techniques for higher quality results is demonstrated.
Combining a portrait with an abstract background to create a unique composition is featured.
Techniques for removing and replacing backgrounds, as well as adding design elements like halos, are presented.
Exporting the composition at a reduced size and using text-to-image methods for further creative input is explained.
A trick for loading the composition into Automatic 1111 for additional creative control is shared.
Using control nets for depth and open pose to refine the composition is demonstrated.
Inpainting and upscaling the final image to achieve the stunning result is the final step in the process.
An accidental deletion of a part of the video is mentioned, offering a live stream as an alternative for the missed content.
The video concludes with a call to action for viewers to like the video and a promise to see them soon in future content.