Creative Exploration - Flux - First Look with Midjourney Man

Purz
2 Aug 2024144:25

TLDRJoin hosts Pers and MJ, the Mid Journey Man, for a creative exploration of AI art generation. They discuss MJ's journey with AI, his background as an engineer, and delve into the capabilities of the new Flux model by Black Forest Labs. The conversation is interspersed with live testing of Flux, sharing their excitement and challenges as they experiment with this cutting-edge technology. They also plan to integrate successful outputs into animations, showcasing the potential of Flux for creative professionals.

Takeaways

  • 😀 The video is a 'Creative Exploration' session featuring a conversation with 'MJ', also known as 'Mid Journey Man', an artist and engineer who creates AI-generated art.
  • 🎨 MJ has been involved with AI art for over two years, starting with 'mid Journey' and has a significant presence on social media platforms like Instagram, Twitter, and YouTube.
  • 🤖 MJ's professional background is in engineering, which influences his approach to tinkering with AI art tools, seeking to understand and manipulate them at a deeper level.
  • 🔧 The discussion highlights the appeal of tools like 'Comfy', which allow users to experiment with node-based visual programming, offering more flexibility than traditional automatic systems.
  • 🌐 The video mentions 'Flux', a new AI model released by 'Black Forest Labs', a company formed by former members of the 'Compis' team from 'Stability AI', known for their work on 'Stable Diffusion'.
  • 🚀 'Flux' is presented as a potentially superior model to 'Mid Journey', sparking excitement for its capabilities and the possibility of testing it during the stream.
  • 📚 The script discusses the importance of understanding licensing terms for AI models, particularly the stipulation against using outputs to train competing models.
  • 💾 Technical challenges arise when attempting to run the 'Flux' model, with the large file size causing potential issues for streaming and memory usage.
  • 🛠️ The video provides a practical walkthrough of troubleshooting and setting up AI models in 'Comfy', including downloading necessary files and adjusting settings.
  • 🎥 There's an emphasis on the creative potential of AI art tools, with plans to experiment with animations and unique visual effects using the outputs from 'Flux'.
  • 🏆 The conversation also celebrates the achievements of the AI art community, with a shoutout to 'Phil', a prominent figure in the field, and 'MJ's' recent win in a project contest.

Q & A

  • What is the title of the video and who are the hosts involved in the 'Creative Exploration'?

    -The title of the video is 'Creative Exploration - Flux - First Look with Midjourney Man'. The hosts involved are Pers and MJ, also known as Mid Journey Man.

  • Where can the audience find Mid Journey Man's work and social media profiles?

    -The audience can find Mid Journey Man's work and social media profiles on Instagram, Twitter, and YouTube, with links provided in the description below the video.

  • How did Pers and MJ meet and what do they have in common?

    -Pers and MJ met at the Comfy Conference in LA. They hit it off and found common ground as kindred spirits and artists on the same wavelength.

  • What is MJ's background and how did he get started with AI art?

    -MJ, whose real name is Julian, is an engineer by trade. He got started with AI art after a friend shared some images made in Mid Journey version two, which piqued his interest and led him to join the beta and start exploring AI art deeply.

  • What does MJ enjoy about working with AI art tools like Comfy UI?

    -MJ enjoys the flexibility and creative control offered by Comfy UI, which allows him to experiment with different elements and create unique visual outputs that reflect his intentions, rather than relying on automatic processes.

  • What is Flux and what is its significance in the AI art community?

    -Flux is a new model released by Black Forest Labs, a company formed by the Compis team that left Stability. It is significant because it offers commercial and open versions of AI models, with the stipulation that outputs cannot be used to create competing models.

  • What are the technical requirements for using the Flux model?

    -The Flux model requires a significant amount of VRAM, with the Dev model using up to 25 gigabytes for inference. It also requires a 10 gigabyte T5 XXL clip and a different VAE model.

  • What is the alternative for users who do not have a high VRAM GPU?

    -For users with less VRAM, there is an fp8 (floating point 8) version of the Flux model provided by Kaji, which allows for lower VRAM usage during inference.

  • How can users who do not have the necessary GPU power access Flux models?

    -Users without the necessary GPU power can access Flux models through the API provided by Black Forest Labs for a token-based fee, which is one cent per image.

  • What is the process of setting up the Flux model in Comfy UI?

    -To set up the Flux model in Comfy UI, users need to download the corresponding workflows and models from the provided links, place them in the appropriate folders within their Comfy UI directory, and update Comfy UI to recognize the new additions.

  • What is the significance of the 'fp8' version of the T5 XXL clip model?

    -The 'fp8' version of the T5 XXL clip model is a smaller, 5-gigabyte version of the original 10-gigabyte clip model, designed for users with less VRAM to still be able to use the model for AI art generation.

Outlines

00:00

🎨 Artistic Collaboration and AI Introduction

The script introduces a collaborative session between two artists, Pers and MJ (Mid Journey Man), discussing their backgrounds, interests in AI art, and the excitement of exploring new AI models like Flux. They highlight the creative process, the joy of finding like-minded artists, and the potential of AI in enhancing their work. MJ shares his journey into AI art, starting with Mid Journey and evolving through various AI platforms, emphasizing his engineering background and how it fuels his passion for AI art experimentation.

05:01

🤖 AI Art Model Exploration: Flux and Technical Details

The conversation delves into the exploration of the new AI art model, Flux, developed by the former Compis team from Stability, now Black Forest Labs. They discuss the commercial and open-source aspects of Flux, its licensing terms, and the potential for both commercial and non-commercial use. The artists express excitement about testing Flux's capabilities and compare it to Mid Journey, discussing the technical specifications, including model size, VRAM usage, and the potential for creating unique animations with the generated images.

10:02

🔧 Technical Setup and Model Installation

The script details the technical setup process for using the Flux model in Comfy UI, including downloading workflows, models, and necessary files like the T5 XXL clip and VAE files. The artists discuss the steps for setting up the environment, updating Comfy UI, and configuring the model settings. They also mention troubleshooting steps, such as managing VRAM limitations and ensuring the correct installation of dependencies.

15:06

🖼️ Testing AI Art Generation with Flux

The artists begin testing the Flux model by running it through Comfy UI, sharing their initial experiences, and discussing the results of their first attempts at generating AI art. They encounter technical challenges, such as running out of VRAM and RAM, and work through solutions like using lower precision models. The conversation is filled with real-time problem-solving, anticipation, and the excitement of exploring new AI capabilities.

20:07

🎭 Experimenting with AI Art Prompts and Techniques

As the artists continue to experiment with Flux, they discuss using various prompts and techniques to generate unique AI art. They share their thoughts on the results, comparing the output quality and aspect ratio handling of Flux with other models. The conversation also touches on the community aspect of AI art, mentioning other artists and the shared excitement in the community about the new model.

25:09

🛠️ Adjusting Technical Settings and Model Comparison

The script focuses on adjusting technical settings for AI art generation, such as changing the diffusion model node to FP8 to reduce memory usage. The artists compare different models, discuss the pros and cons of each, and share their strategies for managing resources during the AI art creation process. They also mention the importance of updating Comfy UI for better workflow management.

30:10

🎨 AI Art Generation Showcase and Discussion

The artists showcase the AI art generated using Flux, expressing their impressions on the quality and style of the images. They discuss the potential for post-processing and the unique aspects of Flux's output compared to other models. The conversation is filled with enthusiasm and a sense of discovery as they explore the capabilities of Flux for creating detailed and intricate art.

35:13

🔄 Exploring AI Art Variations and Styles

The script describes the artists' exploration of various styles and prompts with Flux, generating a diverse range of AI art images. They discuss the versatility of the model in handling different prompts and its ability to create illustrations and other styles. The artists share their excitement about the results and the potential for further experimentation.

40:13

🏆 Recognition and Future AI Art Plans

The conversation includes a moment of recognition for MJ's win in a project contest, highlighting the quality of his AI art piece. The artists discuss their future plans for creating more AI art, including using the generated images as starting points for animations and other projects. They express their excitement about the possibilities opened up by Flux and the ongoing evolution of AI art tools.

45:14

🤝 Collaborative AI Art Session Wrap-up

The script concludes with the artists wrapping up their collaborative AI art session, reflecting on their experiences with Flux and the creative process. They discuss the community aspect of AI art, the importance of sharing work and prompts, and the potential for future collaborations. The artists express their gratitude for the opportunity to explore AI art together and look forward to future sessions.

Mindmap

Keywords

Creative Exploration

This term refers to the process of innovatively investigating and experimenting with new ideas or mediums. In the context of the video, 'Creative Exploration' is the overarching theme, as the hosts delve into the capabilities of an AI model called 'Flux', discussing its potential for artistic creation and manipulation.

Midjourney Man (MJ)

Midjourney Man, also known as MJ, is a nickname for a person who has been deeply involved with AI art since its early stages. The name 'Midjourney Man' suggests someone who is well-versed in the journey of AI art development and contributes significantly to the community. In the video, MJ shares his insights and experiences with AI art tools.

AI Scene

The 'AI Scene' refers to the community and environment surrounding the use of artificial intelligence in creative processes, particularly in art. The script mentions that MJ has been part of the AI scene for over two years, indicating his involvement and expertise in this rapidly evolving field.

Disco Diffusion

Disco Diffusion is a term used to describe a specific AI-based image generation tool that was popular at the time of the video. It is an example of the type of software that the hosts are discussing and using to explore new creative possibilities in their work.

Def Animations

Def Animations likely refer to a style or method of creating animations using AI tools that involve deformation or 'def' techniques. The script suggests that the host is fascinated by these animations and their creative potential.

Comfy

In the script, 'Comfy' seems to refer to a user-friendly AI platform or tool that allows for the tinkering and customization of AI models without the need for extensive coding knowledge. It's a tool that enables users to experiment with AI in a more accessible way.

Flux

Flux is a new AI model introduced by Black Forest Labs, as mentioned in the script. It is a significant topic in the video, as the hosts are excited to explore its capabilities, which are claimed to be as good as, if not better than, Midjourney. The hosts test its performance and discuss its potential impact on AI art.

VRAM

VRAM, or Video Random Access Memory, is a type of memory used by graphics processing units to store image data. In the context of the video, the hosts discuss the VRAM requirements for running the Flux model, indicating that it is a high-resource model that may cause challenges for users with limited hardware capabilities.

Hugging Face

Hugging Face is a company that provides a platform for developers to share and collaborate on AI models. In the script, it is mentioned as the platform where the Flux model is released, indicating it as a hub for AI art enthusiasts and developers to access and experiment with new models.

Animation Techniques

The hosts discuss MJ's approach to animation, which involves intention and deliberate choice rather than random generation. This highlights a key aspect of creative work with AI, where the artist's direction and input can significantly influence the outcome of the generated content.

Highlights

Introduction of 'Flux', a new AI model by Black Forest Labs, the team behind the CompVis model from Stability AI.

Flux is available for commercial use with a unique licensing model that prohibits training on competing models but allows commercial output creation.

The release includes two models: 'Dev' and 'Schnell', with 'Dev' being a 25GB model and 'Schnell' a faster, diffused model.

MJ (Mid Journey Man) shares his journey with AI, starting with mid-journey and his addiction to AI-generated art.

MJ's background as a mechanical engineer influences his approach to using AI tools like Comfy and his preference for visual node-based systems.

Discussion on the creative freedom offered by Comfy, allowing users to experiment with unconventional combinations of elements.

The hosts experiment with running Flux on their systems, facing challenges with VRAM limitations and model loading.

MJ's approach to animation is highlighted, emphasizing intention and deliberate choices over random outcomes.

The potential of Flux for creating high-quality images that can be used for commercial work without ownership restrictions.

Comparison between Flux and other AI models like Mid Journey, discussing the potential of Flux to be superior.

The hosts explore the capabilities of Flux by running various prompts and observing the quality and speed of image generation.

A switch to using the 'Dev' model due to its higher quality output compared to 'Schnell'.

Experimenting with different resolutions and aspect ratios to test Flux's adaptability.

The community's engagement with the live stream, providing prompts and feedback on the AI-generated images.

MJ's success in a recent AI art contest, showcasing the potential of AI in creating award-winning artwork.

The hosts' decision to continue exploring Flux's capabilities in subsequent sessions, indicating the depth of Flux's features.

Final thoughts on Flux's potential impact on the AI art community and the excitement surrounding its release.