BeautifAI Image Upscaler! ft. MidJourneyMan & SuperBeast.ai // Civitai Guest Stream

In this Twitch guest stream, Tyler from Civitai welcomes Julian (MidJourneyMan) and Brad (SuperBeast) to discuss their collaborative work on an open-source image upscaling tool. They share their experiences with AI art and the evolution of upscaling tools, emphasizing the importance of open-source contributions to the community. The workflow they present offers a competitive alternative to paid upscaling services, utilizing Stable Diffusion models and various nodes to enhance image quality without losing original details. The conversation covers the technical aspects of the workflow, including the use of custom nodes, control nets, and the significance of the 'freedom' parameter in determining the level of detail enhancement. They also address practical tips for users, such as experimenting with different models, adjusting settings, and the potential need for post-processing in Photoshop. The stream concludes with a Q&A session and a reminder for viewers to share their upscaled creations on social media, tagging the creators.

Takeaways

• 🎨 Tyler introduces a guest stream featuring Julian (MidJourneyMan) and Brad (SuperBeast.ai), focusing on an open-source image upscaling tool they've developed.
• 🔍 The upscaling tool is designed to compete with paid upscalers and is the result of over a month of collaboration and numerous iterations by Julian and Brad.
• 📈 The workflow includes a 'freedom' parameter, allowing users to control the level of detail enhancement and the degree of deviation from the original image.
• 🚀 The tool utilizes AI art techniques and is particularly useful for those looking to upscale images beyond basic capabilities while maintaining detail.
• ⚙️ The upscaling process involves two passes, with the first pass focusing on a basic upscale and the second pass further enhancing the image resolution.
• 🖼️ The workflow is customizable, with options to adjust settings such as the upscaling factor, denoising levels, and the use of control nets to maintain image structure.
• 💡 Julian and Brad emphasize the importance of experimenting with different settings and models to achieve the desired results.
• 👀 They also discuss the use of custom nodes and the integration of tools like Ultimate Upscale and TT Planet for SDXL in the upscaling process.
• 📉 The stream highlights the potential for artifacts when upscaling, especially when the 'freedom' parameter is set too high, and the need for careful tweaking to avoid unwanted results.
• 🛠️ Editing the original image before upscaling can help improve results, as the upscaling tool can fix certain issues but not add missing details like fingers.
• 🌐 The presenters encourage viewers to share their upscaled images on social media and tag the creators to showcase the tool's capabilities.

Q & A

What is the title of the stream?

The title of the stream is 'BeautifAI Image Upscaler! ft. MidJourneyMan & SuperBeast.ai // Civitai Guest Stream'.

Who are the special guests featured in the stream?

The special guests featured in the stream are Julian, also known as MidJourneyMan, and Brad, also known as SuperBeast.

What is the main topic of discussion in the stream?

The main topic of discussion is an open-source image upscaling tool that the guests, Julian and Brad, have been working on, which they believe can compete with paid upscaling tools.

How long have Julian and Brad been working on the upscaling tool?

Julian and Brad have been working on the upscaling tool for over a month, with around 12 or 13 versions and iterations.

What is the workflow's approach to upscaling images?

The workflow uses a combination of nodes and tools, including Ultimate Upscale and Control Nets, to maintain the structure of the image while upscaling. It also introduces a 'freedom' parameter to control the level of enhancement and detail addition.

What is the significance of the 'freedom' parameter in the upscaling process?

The 'freedom' parameter allows users to control the degree of change and enhancement in the upscaled image. A higher freedom value results in more significant changes and detail addition, while a lower value results in a more conservative upscale that adheres closely to the original image.

How can users find and download the workflow?

Users can find and download the workflow by visiting civii.com and searching for 'beyy', which will provide a link to download the workflow files.

What are the system requirements to run the workflow successfully?

The workflow can run on a system with at least 8GB of VRAM, but it is recommended to have more for better performance. During the stream, the host mentioned that their system was using around 20GB of VRAM.

What are the potential issues that users might encounter when using high freedom values?

Using high freedom values can lead to artifacts in the upscaled image, such as additional heads, eyes, or other unwanted elements. It's important to find a balance where the image enhances while still retaining its original integrity.

How does the workflow handle multiple passes for upscaling?

The workflow supports two passes for upscaling. The first pass provides an initial upscale, and the second pass further enhances the image and increases the resolution. Users can choose to do one or both passes depending on their needs.

What are some tips for using the upscaling tool effectively?

Some tips include editing the original image to remove potential sources of artifacts, experimenting with different SDXL models and Luras for style control, and adjusting the freedom parameter to find the right balance between detail and adherence to the original image.

Outlines

00:00

🎥 Introduction to the Guest Creator Stream

The video script opens with an enthusiastic introduction by Tyler, the host, who welcomes viewers to the civi.tha.com Friday Guest Creator stream. He introduces the special guests, Julian (mid journeyman) and Brad (super beast), both of whom are accomplished artists and have been working on an exciting AI art project. Tyler teases the reveal of an open-source image upscaling tool they've developed, hinting at its potential to compete with paid alternatives.

05:01

🤖 Upscaling Tool Workflow Overview

Julian and Brad provide an overview of their AI art upscaling tool, discussing its development process and the various iterations they have gone through. They mention the use of AI art for two years and the transition from Mid Journey to Comfy UI. The duo outlines the steps to download and implement the tool using Comfy UI, emphasizing the importance of the 'freedom' parameter for controlling the level of image enhancement.

10:02

🔍 Deep Dive into the Upscaling Process

The video offers a detailed look at the upscaling process, explaining the functionality of different nodes and settings within the workflow. It discusses the importance of the 'freedom' parameter in allowing the algorithm to introduce new details and the potential for artifacts at higher settings. The presenters also touch on the use of control nets, the upscaling factor for each pass, and the system's VRAM usage.

15:04

🖼️ Before and After Image Comparison

The script includes a demonstration of the upscaling tool's effectiveness through a before-and-after image comparison. It highlights the tool's ability to clean up pixelization and enhance details, such as the water in an image becoming more realistic. The presenters also discuss the use of filters, sharpening, and color matching to further improve the upscaled images.

20:05

🔧 Customizing and Troubleshooting the Workflow

The video provides tips and tricks for customizing the upscaling workflow and addressing potential issues. It suggests editing the original image to prevent artifacts and experimenting with different SDXL models and Lora prompts. The presenters also recommend testing with different seed values and adjusting the denoising level to balance between preserving the original image and introducing new details.

25:07

📈 Maximizing Image Quality with HDR Effects

The script discusses the use of HDR effects to enhance the visual appeal of upscaled images, allowing for deeper blacks and brighter highlights. It differentiates the HDR effect from simple contrast adjustment and shows how to achieve a more polished look. The presenters also mention the ability to adjust the strength of the HDR effect to suit individual preferences.

30:09

🌟 Showcasing Results and Encouraging Experimentation

The video showcases several examples of upscaled images, demonstrating the potential results viewers can expect from using the workflow. It emphasizes the importance of experimenting with different settings and prompts to achieve desired styles. The presenters also discuss the potential for combining the upscaling tool with image editing software like Photoshop for further refinement.

35:10

📚 Final Tips and Social Media Shoutouts

The video concludes with final tips from the presenters, encouraging viewers to experiment with the workflow and share their results on social media, tagging the creators. It also provides information on where to find the workflow for download and how to use it effectively. The host expresses gratitude to the guest creators and provides a preview of upcoming streams.

Keywords

Image Upscaler

An image upscaler is a software tool or algorithm that increases the resolution of digital images without a significant loss in quality. In the video, the discussion revolves around a new open-source workflow for upscaling images, which is a significant topic as it relates to enhancing image resolution for professional and artistic purposes. The upscaler mentioned is designed to compete with paid tools, indicating its high quality.

MidJourneyMan

MidJourneyMan, whose real name is Julian, is one of the special guests in the video and a significant contributor to the development of the image upscaling workflow. He is recognized for his expertise in AI art and stable diffusion, which are key to the upscaling process discussed. His involvement adds credibility to the workflow's effectiveness.

SuperBeast.ai

SuperBeast.ai, represented by Brad in the video, is another key contributor to the upscaling workflow. Brad's focus on high-resolution art and his experience with various AI tools make him a valuable participant in the development and demonstration of the upscaling process. His contributions are integral to the workflow's ability to enhance details in upscaled images.

Workflow

In the context of the video, a workflow refers to the series of steps or processes involved in using the image upscaling tool. The workflow is detailed and includes various nodes and settings that users can adjust to achieve desired results. It is a core component of the video's narrative as it guides viewers on how to use the upscaling tool.

Stable Diffusion

Stable Diffusion is a term used to describe a type of AI model that is capable of generating detailed images from textual descriptions. In the video, it is mentioned in relation to the upscaling process, suggesting that the workflow may utilize such models to enhance image details during upscaling, which is a critical aspect of achieving high-quality results.

Control Nets

Control Nets are AI components used within the upscaling workflow to help maintain the structural integrity of the image as it is being upscaled. They are essential for preventing artifacts and ensuring that the upscaled image retains its original composition, making them a vital part of the discussed upscaling tool's functionality.

Freedom Value

The Freedom Value is a parameter within the upscaling workflow that allows users to control the degree of change or enhancement applied to the image during the upscaling process. A higher freedom value results in more significant changes and detail enhancement, which is demonstrated in the video through various examples showing the effects of different freedom values.

VRAM Usage

VRAM, or Video RAM, usage refers to the amount of memory used by the graphics processing unit (GPU) during the upscaling process. In the video, the hosts discuss the VRAM requirements for running the workflow effectively, which is crucial information for users to understand the hardware demands of the upscaling tool.

Tiling

Tiling in the context of the video refers to a method used in the upscaling process where an image is divided into smaller parts, or tiles, which are then upscaled individually before being reassembled. This technique is highlighted as a key aspect of the upscaling workflow that contributes to the enhanced resolution and detail of the final image.

Ultimate Upscale

Ultimate Upscale is mentioned as a component or feature within the upscaling workflow that has been around for some time. It is likely a reference to a specific upscaling algorithm or technique that is being integrated into the new workflow to achieve high-quality results, underscoring the combination of existing and new technologies in the process.

Artifacting

Artifacting refers to the unintended visual anomalies or distortions that can occur in upscaled images, such as extra faces or eyes appearing in the picture. The video discusses how the upscaling workflow aims to minimize artifacting through the use of control nets and careful adjustment of the freedom value, which is important for maintaining image quality.

Highlights

Tyler introduces a special guest stream featuring Julian (MidJourneyMan) and Brad (SuperBeast.ai), discussing AI image upscaling tools.

Julian and Brad have been working on an open-source workflow for image upscaling that rivals paid tools.

The upscaling workflow has been in development for over a month with multiple iterations.

The process doesn't just upscale images but enhances details, offering more control to the user.

The workflow is designed to be customizable, allowing users to experiment with different settings and nodes.

Julian and Brad demonstrate the upscaling process with a live example, showing the before and after results.

The upscaling tool uses a 'freedom' parameter that determines how much the image can change and enhance.

Different upscaling models can be used within the workflow for various styles and levels of photorealism.

The workflow includes options for color matching, HDR effects, sharpening, and filters for final image adjustments.

Users are encouraged to share their upscaled images on social media and tag the creators for feedback and visibility.

The workflow is suitable for various uses, from social media to high-resolution prints.

The importance of good initial images and prompts for successful upscaling results is emphasized.

Tips for editing and adjusting the workflow for best results are provided, including the use of seeds for consistency.

The final upscaled image can be significantly larger and of higher quality than the original, with the potential for 4K or higher resolutions.

The workflow is adaptable to work with different AI models and is not limited to a specific upscaling model.

A YouTube tutorial is available to guide users through the workflow installation and usage.

The workflow can be saved as a template and integrated into other projects for a streamlined upscaling process.

The stream concludes with a reminder to follow and tag the creators on social media for community engagement and support.