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

Civitai
8 Apr 202467:31

TLDRIn this engaging Twitch stream, Tyler welcomes guests 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 challenges of upscaling while introducing a workflow that competes with paid tools. The tool incorporates various nodes and allows users to control the level of detail and enhancement through a 'freedom' parameter. The stream also covers best practices for using the tool, including tips on prompts, model selection, and post-processing. The hosts encourage experimentation and provide a YouTube tutorial for users to follow along. The workflow is available for download, and the creators invite users to share their results on social media.

Takeaways

  • 🎨 The stream features a discussion on AI image upscaling tools, focusing on an open-source workflow developed by Julian (MidJourneyMan) and Brad (SuperBeast.ai).
  • 🤖 Julian, who initially used MidJourney, has transitioned to using Stable Diffusion, which is significant as the upscaling tool is designed for Stable Diffusion models.
  • 🚀 The upscaling workflow is designed to be competitive with paid tools and is offered for free, aligning with the open-source ethos of the creators.
  • 🌐 The workflow has been in development for over a month, with numerous iterations to refine the process and enhance image quality.
  • 🔍 The upscaling process involves two passes, with the second pass further enhancing the image and increasing resolution, although it takes longer to process.
  • 📈 A key feature of the workflow is the 'freedom' parameter, which allows users to control the degree of image enhancement and detail addition.
  • 💾 The workflow is resource-intensive, utilizing up to 20GB of VRAM during the upscaling process, indicating the need for a powerful system to run it smoothly.
  • 🛠️ Users are encouraged to experiment with different models, luras (latent representations), and settings to achieve desired styles and effects.
  • 🔧 The workflow includes nodes for post-processing effects such as HDR, sharpening, and grain addition, which can be adjusted or disabled as needed.
  • 🔄 The use of random seeds in the upscaling process helps to prevent 'burn-in' of details and allows for new details to emerge with each pass.
  • ⚙️ The workflow is highly customizable, with options to disable certain nodes and adjust parameters to fit various upscaling needs and personal preferences.

Q & A

  • What is the main topic of the guest stream on civi.tha.com?

    -The main topic of the guest stream is about an AI image upscaler tool developed by Julian, also known as MidJourneyMan, and Brad, also known as SuperBeast.ai. They discuss their workflow, how it competes with paid upscalers, and share their experiences in using it.

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

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

  • What is the significance of the 'freedom' parameter in the upscaling tool?

    -The 'freedom' parameter allows users to control the level of change or enhancement applied to the image during the upscaling process. A higher freedom value results in more significant changes and detail enhancements, while a lower value results in a more conservative upscale that retains the original image characteristics.

  • What are the recommended steps for a user new to the upscaling tool?

    -New users should start by downloading the workflow from civi.tha.com, extracting the files, and then dragging and dropping the JSON and PNG files into the Comfy UI. They should then load their image, input their prompt, and adjust the freedom parameter to their desired level before running the upscaling process.

  • What is the purpose of the 'seam fix mode' in the Ultimate SD upscale node?

    -The 'seam fix mode' helps to denoise the edges of the tile when it is put back together, which can be particularly useful for creating a clean and polished image. However, for testing purposes, users can change it to 'none' to speed up the process significantly.

  • How does the upscaling tool handle different styles of images?

    -The upscaling tool is versatile and can handle different styles of images. However, the results may vary depending on the image style. Users are encouraged to experiment with different SDXL models and prompts to achieve the desired style and level of detail.

  • What are the system requirements to run the upscaling tool?

    -The upscaling tool is resource-intensive and benefits from a powerful GPU. During the stream, the presenters mentioned that their system was using around 20 gigabytes of VRAM, suggesting that a high-end graphics card, such as an RTX 40 series, would be suitable for running the tool.

  • How does the upscaling tool deal with artifacts that may appear during the upscaling process?

    -The tool uses control nets, such as canny and depth, to maintain the structure of the image and counteract artifacts. Additionally, users can adjust the freedom parameter to find a balance that minimizes unwanted artifacts without compromising on the quality of the upscale.

  • What is the role of the HDR effect in the upscaling tool?

    -The HDR effect is used to enhance the visual appeal of the upscaled image by adjusting the highlights and shadows to make them more pronounced. This can make the image 'pop' more, with deeper blacks and brighter highlights, although the strength of the HDR effect can be adjusted to suit individual preferences.

  • Can the upscaling tool be used with other AI models besides the ones mentioned?

    -Yes, the upscaling tool is designed to be flexible and open-ended, allowing users to experiment with different AI models, including various SDXL models and custom LoRAs (Large Optimized Reduced Attention models) to achieve different results.

  • How can users share their results and get feedback on their upscaled images?

    -Users are encouraged to post their upscaled images on social media, tagging the developers MidJourneyMan and SuperBeast.ai, so they can see and share the creations. Additionally, posting images on the civi.tha.com platform with appropriate tags can help others discover and use the upscaling tool.

Outlines

00:00

🎥 Introduction and Guest Announcement

The video begins with an enthusiastic introduction to the Friday guest creator stream hosted by Tyler. He expresses excitement for the viewers to join and introduces the special guests, Julian (mid journeyman) and Brad (super beast), who are both artists with a focus on AI art. Tyler mentions that they will be sharing an open-source workflow for image upscaling, which they believe can compete with paid tools. Julian and Brad have been working on this project for over a month, iterating through multiple versions to refine the process.

05:01

🖼️ Workflow Overview and Download Instructions

Julian and Brad provide an overview of the image upscaling workflow they've developed. They explain that the workflow is designed to enhance images beyond basic upscaling, offering more control over the final product. Julian shares his screen to guide viewers through the process of downloading and installing the workflow from a provided link. They discuss the importance of the freedom parameter, which allows users to control the level of enhancement and detail in the upscaled images.

10:02

🎨 Customizing the Workflow and Editing Tips

The discussion shifts to the customization of the workflow and the importance of editing. Julian and Brad explain how users can adjust the workflow to fit their needs, including changing the upscaling factor and using different models and prompts. They also provide tips on editing the upscaled images to remove artifacts and achieve the desired look. The hosts emphasize the iterative nature of the process and the importance of experimenting with different settings.

15:04

🔧 Understanding Control Nets and Upscaling Factors

Julian and Brad delve deeper into the technical aspects of the workflow, discussing the role of control nets and upscaling factors. They explain how the workflow uses these elements to maintain the structure of the original image while adding details and enhancing resolution. The hosts also talk about the importance of balancing the freedom parameter with the control nets to avoid introducing unwanted artifacts in the upscaled images.

20:05

🌐 Community Interaction and Model Recommendations

The conversation includes interaction with the community, addressing questions and comments from viewers. Julian and Brad offer recommendations for specific upscaling models and discuss the benefits of using different models for different styles. They also talk about the importance of testing and iterating to find the best settings for each image. The hosts encourage viewers to share their results and experiences with the workflow in the community.

25:07

🎯 Final Thoughts and Stream Wrap-up

As the stream comes to an end, Julian and Brad share their final thoughts on the workflow and the potential it offers for creatives. They reiterate the importance of experimentation and customization, and encourage viewers to continue exploring the possibilities of the workflow. Tyler thanks the guests for their contributions and provides social media shoutouts. He also teases the next guest creator stream and encourages viewers to join and learn from the upcoming presentations.

Mindmap

Keywords

💡Image Upscaler

An image upscaler is a software tool used to increase the resolution of digital images without losing quality. In the context of the video, it refers to the main subject of discussion, which is a new open-source tool developed by the guests to enhance and upscale images, potentially rivaling paid alternatives.

💡MidJourneyMan

MidJourneyMan, whose real name is Julian, is one of the guest creators featured in the video. He is known for his work with AI art and stable diffusion models. He plays a significant role in demonstrating the image upscaler tool and shares insights based on his experience.

💡SuperBeast.ai

SuperBeast.ai, run by Brad, is another entity represented in the video. Brad is involved in the development of the image upscaler and contributes to the discussion by sharing his expertise in AI art, particularly in the context of anthro beastly art.

💡Stable Diffusion

Stable Diffusion is a term referring to a type of AI model used for generating images from textual descriptions. It is significant in the video as the developers have transitioned from using MidJourney to focusing on stable diffusion models, which are utilized in the image upscaler workflow.

💡Open Source

Open source indicates that the software's source code is available to the public, allowing anyone to view, use, modify, and distribute the software. The video emphasizes the open-source nature of the image upscaler, suggesting that it can be freely accessed, modified, and improved by the community.

💡Workflow

In the video, a workflow refers to the series of steps or processes involved in using the image upscaler tool. The workflow is detailed and includes various nodes and settings that users can adjust to upscale their images in a desired manner.

💡Tiling

Tiling in the context of the video is a technique used in the image upscaler to break down the image into smaller parts (tiles), process them individually, and then reassemble them to form a higher resolution image. It is a key part of the upscaler's functionality.

💡Freedom Value

The freedom value is a parameter within the image upscaler that controls the degree of change and enhancement applied to the image. A higher freedom value allows for more significant alterations and detail addition, while a lower value results in a more conservative upscale that adheres closely to the original image.

💡Control Nets

Control Nets are algorithms used within the image upscaler to help maintain the structural integrity of the image during the upscaling process. They are crucial for preventing artifacts and ensuring that the upscaled image retains the essential characteristics of the original.

💡HDR Effect

HDR, or High Dynamic Range, effect is a post-processing tool mentioned in the video that enhances the contrast and color range of the upscaled image. It is used to make the image 'pop' more by adjusting the highlights and shadows, resulting in a more visually striking final product.

💡Civitai

Civitai is a platform where the discussed image upscaler workflow is made available for download and use. It serves as a community hub where creators like MidJourneyMan and SuperBeast.ai share their work and where users can access and utilize these shared resources.

Highlights

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

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 many iterations to achieve optimal results.

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

The workflow is designed to be user-friendly, allowing for easy installation and use of the custom nodes.

Users can experiment with different SDXL models and prompts to achieve desired styles and effects.

The upscaling process involves two passes, with the second pass further enhancing the image resolution.

The workflow is adaptable and can be modified by users to improve it or tailor it to their needs.

Julian and Brad discuss the importance of the 'freedom' parameter in controlling the level of image enhancement.

The upscaling tools can handle images with repetitive details like the skulls example, adding new details effectively.

The stream demonstrates the use of the workflow with a live example, showing the before and after results.

The workflow includes options for color matching, HDR effect, sharpening, and grain addition to fine-tune the final image.

Julian emphasizes the potential for artifacts when using high 'freedom' values and the need for careful parameter adjustment.

Brad shares tips for editing the initial image to prevent artifacts and for achieving a more photo-realistic look.

The presenters discuss the use of different seeds in upscaling passes to prevent detail burn-in and over-enhancement.

The workflow is showcased as a flexible tool that can be integrated into other image processing workflows for enhanced results.

The stream concludes with a Q&A session where the presenters answer questions about the upscaling process and the workflow's capabilities.

Create Stunning AI Art Instantly

No login. No cost. Just incredible AI images !

Create Stunning AI Art InstantlyGenerate Now