VIDEO
HIGHLIGHT
Log InSign up
MidJourney In Trouble? Open-Source and FREE AI Generator SDXL Is Stunning (Comparison)
SummaryTranscriptChat

MidJourney In Trouble? Open-Source and FREE AI Generator SDXL Is Stunning (Comparison)

Introduction to Stable Diffusion XL

In this section, the speaker introduces Stable Diffusion XL, an AI image generation model that produces high-quality images. The speaker discusses its capabilities and highlights its open-source nature.

Features of Stable Diffusion XL

  • Stable Diffusion XL is an AI image generation model that can create stunning and realistic images.
  • It is similar in quality to mid-journey AI-generated images.
  • The model is free to use and offers nearly unlimited possibilities.
  • It has undergone significant improvements compared to its predecessor, resulting in enhanced image quality and composition detail.

Examples of Improved Image Quality

  • A comparison between the beta version and the new release showcases the significant improvement in image detail and color.
  • : Comparison of two versions with the same prompts, highlighting the increased detail and color in the new release.
  • : Comparison of a wolf image, demonstrating improved details and realism in the new release.
  • : Comparison of an aesthetic hand holding a coffee cup, showcasing flawless details in the new release.

Expanded Functionalities

  • Stable Diffusion XL 0.9 offers a range of functionalities beyond basic text prompting image generation.
  • It includes features like image painting, which replaces portions of an image with generative art, and out painting, which seamlessly extends existing images.

Advancements in Composition for SDXL 0.9

This section focuses on the advancements made in composition for Stable Diffusion XL (SDXL) 0.9. The speaker explains how parameter count plays a crucial role in improving composition quality.

Parameter Count Increase

  • SDXL 0.9 has a significant increase in parameter count compared to the beta version.
  • It boasts a base model with 3.5 billion parameters and an Ensemble Pipeline with 6.6 billion parameters.
  • The use of two clip models, including one of the largest open clip models trained to date, enhances processing power and results in upgraded realism, depth, and resolution.

System Requirements and Availability

This section provides information about the system requirements for using Stable Diffusion XL 0.9 and its availability for research purposes.

System Requirements

  • Stable Diffusion XL 0.9 can run on a modern consumer GPU.
  • Required operating systems include Windows 10 or 11 or a Linux operating system.
  • A minimum of 16 gigabytes of RAM is needed (not VRAM).
  • An Nvidia GeForce RTX 20 graphics card with at least 8 gigabytes of VRAM is recommended.

Availability and Future Releases

  • Stable Diffusion XL can be used through Clip Drop and the API in Dream Studio.
  • SDXL 0.9 will be provided for research purposes only during a limited period to collect feedback before its general open release.
  • The code is already available on Stability AI's GitHub page for users to download and experiment with.

Conclusion and Future Release

In this final section, the speaker concludes by discussing the availability of Stable Diffusion XL's code and hints at future releases.

Code Availability

  • The code for Stable Diffusion XL is available on Stability AI's GitHub page for users to download and explore immediately.

Future Release - SDXL 1.0

  • Stable Diffusion XL 1.0 is targeted for a full open release in mid-July.

Overview of AI features in the free version

The speaker discusses the various AI features available in the free version of the software, including background removal, picture cleanup, relighting, and image upscaling.

AI Features in the Free Version

  • The free version offers several AI features such as background removal, picture cleanup, relighting, and image upscaling.
  • These features enhance the functionality of the software and provide users with a range of options for editing and improving their images.

Comparing images generated by different models

The speaker compares images generated by different models using prompts from a website called Mid Journey. They discuss the quality and artistic value of these images.

Image Comparison with Different Models

  • The speaker tests out a prompt from Mid Journey on both Clipdrop and SDXL.
  • While the images generated by both models do not resemble those found on Mid Journey, they are still visually appealing and artistic.
  • The speaker highlights that SDXL produces even more artistic results compared to what was found on Mid Journey.

Recreating a fingerprint prompt using SDXL

The speaker attempts to recreate a fingerprint prompt using SDXL. They explore different color variations to achieve desired results.

Recreating Fingerprint Prompt with SDXL

  • The speaker uses a broad and generic fingerprint prompt to test SDXL's capabilities.
  • Although the resulting images are black and white, they are visually pleasing.
  • By introducing blue and red colors into the prompt, they are able to achieve results closer to what was found on Mid Journey.

Generating realistic lion image with SDXL

The speaker tests SDXL's ability to generate a realistic image of a lion. They compare the generated images with their expectations.

Generating Realistic Lion Image with SDXL

  • The speaker expresses that the generated lion image is more artistic than realistic, which was not their initial expectation.
  • However, they appreciate the stunning level of detail and realism in the images produced by SDXL.

Exploring additional features of SDXL

The speaker demonstrates and explores additional features of SDXL, including background removal, imperfection cleanup, relighting, and image enhancement.

Additional Features of SDXL

  • The speaker showcases the "relight" feature in SDXL, which allows users to position lights on a face and adjust lighting parameters in real-time.
  • They demonstrate how changing light intensity, radius, and distance affects the appearance of the face.
  • The speaker emphasizes their excitement about the progress made by this model and looks forward to future updates.

Conclusion and anticipation for future updates

The speaker concludes by expressing their enthusiasm for the software's progress and upcoming 1.0 version release. They also mention that it is currently free and open-source but has a daily limit on usage.

Conclusion and Future Updates

  • The speaker expresses excitement about the substantial progress made by this model.
  • They mention that while there is currently a daily limit on usage (400 images per day), it will soon be possible to run the software on one's own computer without any limitations.
  • The speaker acknowledges that Mid Journey has their work cut out for them due to the impressive capabilities of SDXL.
Channel: Matthew Berman
Video description

In this video, we look at the new version of Stable Diffusion, SDXL 0.9, which is a huge leap forward in quality for open-source AI art generators. This puts it on par with MidJourney, but it's free, nearly unlimited, and completely open-source. Plus, it can be run on most modern computers locally. I do a direct comparison, what do you think about the new SDXL? Enjoy :) Join My Newsletter for Regular AI Updates 👇🏼 https://forwardfuture.ai/ My Links 🔗 👉🏻 Subscribe: https://www.youtube.com/@matthew_berman 👉🏻 Twitter: https://twitter.com/matthewberman 👉🏻 Discord: https://discord.gg/xxysSXBxFW 👉🏻 Patreon: https://patreon.com/MatthewBerman Media/Sponsorship Inquiries 📈 https://bit.ly/44TC45V Links: Blog Post - https://stability-ai.squarespace.com/blog/sdxl-09-stable-diffusion Github Repo - https://github.com/Stability-AI/generative-models Clipdrop - https://clipdrop.co/