How to improve 3D people in your renders using AI (in Stable Diffusion) - Tutorial

In this tutorial, the Digital Bunch demonstrates how to enhance 3D people in renders using Stable Diffusion, an open-source AI tool. They discuss the importance of staying updated with AI advancements in the creative industry and provide a step-by-step guide on using Stable Diffusion. The process includes installing the software, using the web interface, cropping and selecting areas of interest in images, and choosing the right model for editing. They explain the use of positive and negative prompts to guide the AI, and detail the settings for optimal results, such as resolution, batch size, and noise strength. The tutorial concludes with tips on achieving more realistic results by selecting different parts of the image and emphasizes that while AI can be a powerful tool, it's not perfect and can sometimes produce unexpected results. The creators encourage viewers to share their experiences and outcomes with Stable Diffusion.

Takeaways

• 🎨 **Stable Diffusion Introduction**: Stable Diffusion is an open-source, deep learning text-to-image model that can improve 3D people in renders using AI.
• 📈 **Feedback and Demand**: The community has shown great interest in AI's role in enhancing creative work, prompting the creation of this tutorial.
• 🔍 **Staying Updated**: It's crucial to keep up with AI advancements as they can significantly impact the creative industry.
• 🚀 **Getting Started**: Before using Stable Diffusion, it needs to be installed, and the web interface (Automatic1111) offers a variety of features and options.
• 🖼️ **Image Processing Limitations**: Stable Diffusion currently does not process large images, requiring users to crop and focus on specific areas.
• 🖌️ **Selecting the Model**: For editing people in images, a model specialized in faces and people, like Realistic Vision, is recommended.
• ✍️ **Crafting the Prompt**: Using a positive and negative prompt helps guide the AI to generate desired results while avoiding undesired ones.
• 🔍 **Resolution and Settings**: An optimal resolution of 768 pixels and batch size of four are suggested for generating images with good quality and detail.
• 🔧 **Noising Strength**: A noising strength between 25 to 45 helps in achieving a balance between realism and change from the original image.
• ⏱️ **Processing Time**: Generating images with Stable Diffusion is computed locally and takes about 1 minute on a 4070 TI card.
• 📝 **Post-Processing**: Pasting the AI-generated image back into the visualization software can yield a more realistic final render.
• 🤖 **AI Limitations**: While Stable Diffusion is powerful, it can sometimes produce strange or unrealistic results, especially with higher noising values.

Q & A

What is Stable Diffusion and when was it released?

Stable Diffusion is an open-source software project that uses deep learning for text-to-image modeling. It was released in December 2022.

How can users access the Stable Diffusion interface?

Users can access the Stable Diffusion interface by installing the software and using the provided desktop shortcut, which opens the platform in a web browser. The URL can be copied from the script and bookmarked for easy access.

What is the limitation of Stable Diffusion when it comes to image processing?

Stable Diffusion currently does not process large images. Users need to crop the part of the image they are most interested in and save it as a separate file for processing.

How does the selection of models in Stable Diffusion work?

Each model in Stable Diffusion is trained to perform a specific task. For editing elements like faces and people, a model specialized in those areas, such as Realistic Vision, should be selected.

What is the purpose of positive and negative prompts in Stable Diffusion?

Positive prompts define the desired outcome, while negative prompts specify the results to avoid. They should be kept simple and clear, often including adjectives that describe the desired quality and characteristics.

What are the optimal settings for the Stable Diffusion model when working with faces and people?

The optimal settings include using a resolution of 768 pixels, setting the batch size to four for generating four different images, and adjusting the denoising strength to a value between 25 to 45 for a balance between realism and change from the original image.

How long does it typically take for Stable Diffusion to generate an image on a 4070 TI card?

It usually takes about 1 minute to generate an image on a 4070 TI card, with the computation being done locally.

What are the potential issues with using Stable Diffusion for image processing?

While Stable Diffusion can produce realistic results, it can also generate artifacts or weird outcomes, especially if the denoising value is set too high. It's also worth noting that AI can sometimes hallucinate, leading to unexpected or unrealistic results.

Can Stable Diffusion be used to improve 3D models that have already been generated?

Yes, Stable Diffusion can be used to further refine and improve 3D models that it has already generated, although the results can vary.

What is the recommended approach for achieving better results with faces in Stable Diffusion?

For better results, it is recommended to select and process the face and body separately in Stable Diffusion, as this can lead to more photorealistic outcomes.

How can users share their results and experiences with Stable Diffusion?

Users are encouraged to share their results and experiences by commenting on the tutorial and providing feedback. This helps the community to learn from each other and improve their use of the tool.

What are the future prospects of using AI like Stable Diffusion in creative industries?

The use of AI in creative industries was once thought to be untouchable by automation, but with the advent of tools like Stable Diffusion, this assumption is changing. The industry needs to stay on top of these advancements to leverage their potential in creative work.

Outlines

00:00

🎨 Introduction to Stable Diffusion in Art Projects

The video begins with the host introducing themselves and their team, the Digital Bunch. They discuss their recent experiments with stable diffusion and AI, which garnered positive feedback and prompted requests for a tutorial. The host emphasizes the importance of staying updated with evolving tools, noting the mixed results from their tests. They highlight the significant progress made in AI's role in the creative industry, which was once thought to be immune to automation. The tutorial covers the basics of using stable diffusion, starting with installation instructions and navigating the web interface. The process involves opening an image in Photoshop, cropping the desired section, and using the web UI to edit specific elements. Different models are mentioned for various tasks, such as 'Realistic Vision' for faces and 'Photon' for vegetation and environments.

05:01

🖼️ Using Stable Diffusion for Image Editing

The tutorial continues with practical steps on how to use stable diffusion for image editing. It covers how to input prompts, differentiate between positive and negative prompts, and adjust settings like resolution and batch size. The host explains the importance of the denoising strength setting, which controls the difference between the original and the newly generated image. They share their experiences with generating images, mentioning that the process usually takes about a minute on their hardware. The video also touches on the possibility of selecting and editing different parts of an image, such as the face and body, for better results. Before concluding, the host discusses the potential of stable diffusion for tweaking clothes and improving generated images, while acknowledging the tool's limitations and occasional unexpected results. They invite viewers to share their experiences and request further demonstrations or tests, expressing enthusiasm for research and development in the field.

Keywords

💡Stable Diffusion

Stable Diffusion is an open-source software project that utilizes deep learning to generate images from text descriptions. It is a rapidly evolving tool in the field of AI, which has garnered significant interest due to its potential to assist artists and designers. In the video, it is used to enhance 3D people in renders, showcasing how AI can be integrated into creative processes to achieve more realistic results.

💡Digital Bunch

Digital Bunch refers to the group or team that the speaker is a part of. They are likely involved in digital art, design, or some form of technology that intersects with creative work. In the context of the video, they are experimenting with Stable Diffusion to improve their projects, indicating their role as early adopters and innovators in their field.

💡Artificial Intelligence (AI)

Artificial Intelligence, often abbreviated as AI, refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. In the video, AI is central to the process of improving 3D renders through Stable Diffusion, as it enables the software to understand and generate more realistic human figures.

💡Deep Learning

Deep Learning is a subset of machine learning that involves the use of artificial neural networks to analyze various factors of data. In the context of the video, Stable Diffusion employs deep learning algorithms to convert textual prompts into visual images, which is crucial for the task of enhancing 3D people in renders.

💡Web Interface

A web interface in this context refers to the online platform or application through which users can interact with Stable Diffusion. The speaker mentions that it can be a bit confusing at first, but it offers many features and options for users to manipulate and generate images as desired.

💡Photoshop

Photoshop is a widely used software for image editing and manipulation, developed by Adobe. In the video, it is used to crop and prepare images for processing by Stable Diffusion, highlighting its role as a complementary tool in the image creation process.

💡Cropping

Cropping in image editing involves cutting out a portion of the image to focus on the subject of interest. The video mentions that Stable Diffusion does not process large images, so users need to crop the part of the image they are most interested in and save it separately before using it with the software.

💡Model

In the context of AI and machine learning, a model refers to a system that has been trained to perform specific tasks, such as generating images. The video discusses selecting a model specialized in faces and people, called 'Realistic Vision,' for the purpose of enhancing 3D people in renders.

💡Prompt

A prompt in the context of AI image generation is a text description that guides the AI in creating an image. It can include both positive (desired outcomes) and negative (unwanted outcomes) elements. The video emphasizes the importance of keeping prompts simple and clear to achieve the best results.

💡Noising Strength

Noising strength is a parameter in AI image generation that determines the level of difference between the generated image and the original. A higher value results in a more significant change. In the video, it is suggested to set this value between 25 to 45 for a balance between realism and change.

💡Resolution

Resolution refers to the amount of detail an image has, typically measured in pixels. The video specifies an optimal resolution of 768 pixels for the Stable Diffusion model to work effectively, noting that larger resolutions may not improve quality and smaller ones may lack detail.

💡Batch Size

Batch size in AI image generation is the number of images the system generates at one time. The video mentions setting the batch size to four, which means that Stable Diffusion will generate four different images for the user to choose from, balancing processing time with options for selection.

Highlights

Stable Diffusion is an open-source software project that uses AI to improve 3D people in renders.

The tutorial provides a step-by-step guide on using Stable Diffusion for enhancing renders.

Stable Diffusion is a deep learning text-to-image model released in December 2022.

To use Stable Diffusion, one must first install the software and use the web interface for interaction.

The web interface, Automatic 1111, can be confusing but offers many features and options.

For optimizing images, one should crop the area of interest and save it as a separate file due to the current limitations on large image processing.

Selecting the right model is crucial; for faces and people, the Realistic Vision model is recommended.

The Photon model is effective for realistic vegetation and environments.

Prompts should be simple, clear, and include both positive and negative aspects to guide the AI.

Defining the element to change and describing lighting and lens settings can refine the AI's output.

Negative prompts help the AI avoid undesired results, such as cartoon or anime styles.

Settings like masked options, resolution, and batch size are important for controlling the AI's output.

The denoising strength setting determines how different the new image will be from the original.

Stable Diffusion can generate multiple images, allowing users to choose the best result.

The tool is adept at tweaking clothes and can sometimes produce more realistic results than 3D models.

Fixing people already generated by Stable Diffusion can lead to further improvements.

AI tools like Stable Diffusion can sometimes hallucinate, leading to unexpected or humorous results.

The tutorial encourages users to share their experiences and outcomes with Stable Diffusion.

The presenter expresses excitement about the potential of AI in the creative industry and invites further exploration.