BeautifAI Image Upscaler! ft. MidJourneyMan & SuperBeast.ai // Civitai Guest Stream
TLDRIn this engaging Twitch stream, Tyler is joined by special guests Julian (MidJourneyMan) and Brad (SuperBeast.ai) to discuss the latest in AI image upscaling. They introduce an open-source workflow designed to compete with paid tools, offering a cost-effective solution for enhancing image quality. The conversation covers the importance of transitions in creating a professional feel, the challenges of working with open-source tools, and the iterative process of refining the upscaling tool. The guests share their expertise in AI art, their experiences with various upscaling tools, and provide a step-by-step guide on how to use the new workflow. They also discuss the technical aspects, such as VRM usage and the potential for customization, and encourage viewers to experiment with the tool and share their results. The stream concludes with a Q&A session and a reminder to follow them on social media for more insights into AI and digital art.
Takeaways
- π¨ The stream is about showcasing an AI image upscaler tool developed by MidJourneyMan and SuperBeast.ai, which can compete with paid upscalers.
- π€ MidJourneyMan, initially known for his work with MidJourney, has shifted focus to Stable Diffusion and has been working on AI art for nearly two years.
- π SuperBeast.ai, run by Brad, specializes in anthro beastly art and has been experimenting with various tools to achieve higher image resolutions.
- π The upscaler workflow has been in development for over a month with numerous iterations to refine the process and enhance image details.
- πΎ The workflow is open source and can be downloaded from Civitai, allowing users to customize and improve it based on their needs.
- πΌοΈ The tool includes a 'freedom' parameter that controls the level of image enhancement and detail addition, with higher values allowing more changes.
- π§ Users can experiment with different SDXL models and control nets within the workflow to achieve desired styles and effects.
- π The workflow is designed for two passes, with the first pass focusing on upscaling and the second pass on further enhancement.
- π The use of control nets like Canny edge detection and depth mapping helps maintain the original image structure during the upscaling process.
- π The final output can be significantly larger, with the second pass potentially resulting in a 4K image from an initial 2K source.
- βοΈ The workflow is resource-intensive, with VRM usage reaching up to 20GB during the stream, suggesting a need for powerful hardware for optimal performance.
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 and plan to share with the viewers.
What is the purpose of the upscaling tool they are discussing?
-The purpose of the upscaling tool is to enhance and upscale images to a higher resolution without losing quality, potentially competing with paid upscaling tools.
How long have Julian and Brad been working on the upscaling workflow?
-Julian and Brad have been working on the upscaling workflow 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 and enhancement applied to the image during the upscaling process, with higher values resulting in more significant changes and detail enhancements.
What are the recommended system requirements to run the upscaling workflow?
-The workflow can run on systems with at least 8GB of VRAM, but for better performance, 20GB or more is recommended, as mentioned by the streamers who were running it on a system with 24GB VRAM.
How can viewers access the upscaling workflow?
-Viewers can access the upscaling workflow by visiting civii.com, searching for 'beyy', and downloading the provided link.
What are the potential issues that may arise when using a high 'freedom' parameter?
-Using a high 'freedom' parameter may result in the creation of artifacts, such as extra faces or eyes, and can potentially deviate significantly from the original image.
What is the role of the 'seam fix mode' in the upscaling process?
-The 'seam fix mode' helps to denoise the edges of the tile when it is put back together, resulting in a cleaner and more polished image.
How can users customize the upscaling tool to their preferences?
-Users can customize the upscaling tool by adjusting the 'freedom' parameter, experimenting with different SDXL models, and using Luras to push the style in the desired direction.
Outlines
π₯ Introduction to the Guest Creator Stream
The video begins with an exciting 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 in the AI art community. Tyler teases an open-source image upscaling tool they've been working on, which they claim can compete with paid alternatives. The guests share a bit about their backgrounds and the process they've developed for upscaling images.
π Upscaling Workflow Overview and Download Instructions
Julian and Brad provide a detailed walkthrough of their upscaling workflow, explaining the process from downloading the workflow files to implementing them in Confy UI. They discuss the importance of the 'freedom' parameter, which allows users to control the level of image enhancement. The paragraph also covers the technical aspects of the workflow, including the use of control nets and the tiling process for upscaling.
π Closer Look at the Upscaling Process and Custom Nodes
The third paragraph delves deeper into the upscaling process, highlighting the before and after results with a slider for comparison. It emphasizes the use of custom nodes created by Brad and the ability to tweak settings for desired outcomes. The paragraph also discusses the use of different SDXL models and the importance of experimenting with various settings to achieve the best results.
πΌοΈ Exploring the Effects of Freedom Values on Upscaling
This section focuses on the impact of the 'freedom' parameter on the upscaling output. It demonstrates how higher freedom values can lead to more significant changes and detail enhancement in the upscaled images. The paragraph also touches on the potential for artifacts when pushing the freedom parameter too high and the need for post-processing to clean up images.
π οΈ Tips and Tricks for Using the Upscaling Workflow
The fifth paragraph offers practical tips for using the upscaling workflow effectively. It suggests editing the initial image to avoid artifacts and emphasizes the importance of experimenting with different models and seeds. The paragraph also discusses the use of the seam fix mode for cleaner images and the option to randomize seeds for varied results in successive passes.
π Understanding the Impact of Different Freedom Levels
The sixth paragraph illustrates the effects of different freedom levels on the upscaling process. It provides visual examples of how varying the freedom parameter can result in minimal to significant changes in the image. The discussion also covers the use of control nets and the importance of maintaining a balance between upscaling and detail retention.
π Showcasing Final Upscaling Results and Encouraging Experimentation
The final paragraph showcases the final results of the upscaling process, highlighting the improved quality and detail of the images. It encourages viewers to experiment with the workflow and share their creations. The paragraph ends with a reminder of the importance of good input for better output and a call to action for viewers to engage with the content and creators on social media.
π£ Closing Remarks and Upcoming Stream Preview
In the concluding paragraph, the host thanks the guests for their contribution and provides a preview of the next week's guest creator stream featuring Spence, a technical researcher and artist known for creating the ZeroScope model. The host also reminds viewers to follow the channel on social media and provides information on where to find the workflow and how to share their upscaled images.
Mindmap
Keywords
π‘Image Upscaler
π‘MidJourneyMan
π‘SuperBeast.ai
π‘Stable Diffusion
π‘Open Source
π‘Workflow
π‘Upscaling Tools
π‘Civitai
π‘Freedom Value
π‘Tile Upscaling
π‘Control Nets
Highlights
Tyler introduces the Civitai Friday guest Creator stream featuring Julian (MidJourneyMan) and Brad (SuperBeast.ai), discussing their work on an open-source image upscaling tool.
The upscaling tool is designed to compete with paid upscalers without charging users, emphasizing the open-source nature of the project.
Julian and Brad have been working on the upscaling workflow for over a month, iterating through 12 to 13 versions.
The workflow incorporates AI art techniques and leverages tools like ControlNet to maintain image structure during upscaling.
The upscaling process involves two passes, with the second pass further enhancing image resolution and details.
The freedom parameter in the workflow allows users to control the level of detail enhancement and image alteration.
The upscaling tool includes a comparison feature that lets users see before and after results with a slider.
Julian and Brad discuss the importance of using the right SDXL models and Lora prompts to achieve desired styles and details.
The workflow is designed to be customizable, allowing users to experiment with different settings and nodes.
System requirements for running the workflow include a VRM usage of around 20GB for a 4K upscaling process.
The stream demonstrates the use of the upscaling tool on various images, showcasing the differences between pass one and pass two results.
Julian and Brad share tips for editing and post-processing upscaled images in Photoshop for better results.
The workflow is adaptable for different use cases, such as upscaling for social media or production-quality images.
The presenters discuss the potential for using the upscaling tool with various SDXL models to achieve different artistic styles.
The stream includes a Q&A session where viewers can ask questions about the upscaling tool and its usage.
The workflow is available for download on Civitai, and users are encouraged to share their results and tag the creators on social media.
Tyler thanks Julian and Brad for their contribution to the open-source community and their dedication to refining the upscaling tool.