The Free & Uncensored Version of MidJourney! (FLUX.1)

Matt Wolfe
6 Aug 202418:24

TLDRThis video explores Flux.1, a new AI image-generating tool by Black Forest Labs, which rivals MidJourney in quality and versatility. The team behind Flux.1 developed key AI models like Stable Diffusion and VQ-GAN. Flux.1 offers three models with varying power and cost, with the open-source Flux.1 Schnell suitable for local development and personal use. The video demonstrates Flux.1's capabilities through various image prompts, highlighting its strengths in realism and text generation, while comparing it to MidJourney and DALL-E 3. The potential of Flux.1 as an uncensored and open-source tool is also discussed, along with its upcoming text-to-video model capabilities.

Takeaways

  • 🌟 Black Forest Labs has released a new AI image-generating tool called Flux.1, which is being compared to MidJourney.
  • 🛠️ Flux.1 was developed by many of the same team members who created Stable Diffusion, known for innovations like VQ-GAN and Latent Diffusion.
  • 🔑 There are three models of Flux.1, each with varying power and cost: Flux.1 Schnell (fastest, open-source), Flux.1 Dev (middle, non-commercial use), and Flux.1 Pro (top-performing, enterprise solutions).
  • 📜 Flux.1 Schnell is available under the Apache 2.0 license, allowing for commercial and non-commercial use of the generated images.
  • 🌐 Several websites have integrated Flux.1, offering free use of the models, including the platform Glyph, which allows users to build their own workflows.
  • 🎨 Flux.1 is not as strong in generating illustrations compared to other models like MidJourney, which excels in this area.
  • 🏆 Flux.1 is praised for its realism and quality of output, being on par with MidJourney when using the right prompts.
  • 📝 Flux.1 is particularly good at handling text within images, making it ideal for creating logos, Snapchat selfies, and memes.
  • 🔍 Flux.1 is designed with uncensored capabilities, allowing for more creative freedom, although it does have an NSFW filter.
  • 🐉 Flux.1 shows promise in prompt adherence, capturing many elements from complex prompts, although it may not outperform models like DALL-E 3 in this regard.
  • 🎥 Flux.1 is set to expand into text-to-video capabilities, offering an open-source alternative to platforms like Lumen Dream Machine and Runway Gen 3 Sora.

Q & A

  • What is the name of the new AI image generating tool discussed in the video?

    -The new AI image generating tool discussed in the video is called FLUX.1, developed by Black Forest Labs.

  • Which team members were involved in the development of FLUX.1 and what are some of their previous innovations?

    -The development of FLUX.1 was led by team members who helped build Stable Diffusion. Their innovations include creating VQ-GAN, Latent Diffusion, and models for image and AI video generation like Stable Diffusion XL, Stable Video Diffusion, and Rectified Flow Transformers.

  • What are the three models of FLUX.1 and how do they differ in terms of power and cost?

    -The three models of FLUX.1 are Schnell, Dev, and Pro. Schnell is the fastest model for local development and personal use, Dev is more efficient and prompt adherent than Schnell, and Pro is the top-line model offering state-of-the-art performance designed for enterprise solutions.

  • What is the licensing situation for the FLUX.1 Schnell model?

    -The FLUX.1 Schnell model is openly available under the Apache 2.0 license, making it open source. Tools created using Schnell and images generated with it can be used both non-commercially and commercially.

  • How can one access and use the FLUX.1 models for free?

    -One can access and use the FLUX.1 models for free through websites that have integrated it, such as Black Forest Labs on Hugging Face, and the platform Glyph, which allows building custom workflows and generating images using the Pro model for free.

  • What is the main difference between the FLUX.1 Dev model and the Schnell model in terms of commercial use?

    -The FLUX.1 Dev model can be used for non-commercial applications, meaning you cannot create a tool using this model and then sell access to that tool, unlike the Schnell model which allows commercial use.

  • What type of images is FLUX.1 particularly good at generating according to the video?

    -FLUX.1 is particularly good at generating realistic images and images that involve text, such as logos, Snapchat selfies, and memes.

  • What are some of the limitations of FLUX.1 when compared to other AI image generating tools like Mid Journey?

    -FLUX.1 may not be as good at generating illustrations, oil paintings, and watercolor paintings as compared to Mid Journey, which seems to produce more authentic-looking art styles in these categories.

  • How does FLUX.1 handle the generation of copyrighted images or images of existing IPs?

    -While the current version of FLUX.1 does not allow for the generation of not safe for work (NSFW) content, it can theoretically generate copyrighted images and existing IPs, although this may change as the open-source model evolves.

  • What is the potential future development for the FLUX.1 model as mentioned in the video?

    -The potential future development for FLUX.1 includes the expansion to a text-to-video model, providing an open-source option to generate video content similar to tools like Lumen Dream Machine and Runway Gen 3 Sora.

  • What are some tips for improving prompts when using FLUX.1?

    -To improve prompts with FLUX.1, one can use an AI method to automatically enhance the prompt, be as detailed as possible, and use descriptive language to push the boundaries of creativity, as demonstrated by users like fler on X.

Outlines

00:00

🚀 Introduction to Flux One AI Image Generator

The video script introduces Flux One, a new AI image-generating tool developed by Black Forest Labs, a team with a strong background in AI image and video generation technologies. Flux One is positioned as a competitor to mid-journey models, with claims of superior performance in certain areas. The script provides an overview of three models: Flux One Schnell, designed for local development and open-source use; Flux One Dev, a middle-tier model for non-commercial applications; and Flux One Pro, the top-tier model for enterprise solutions. The video aims to explore these models, their capabilities, and how they compare to existing tools in the market.

05:02

🎨 Exploring Flux One's Capabilities and Limitations

This paragraph delves into the testing and experimentation with Flux One, highlighting its strengths in generating realistic images and handling text within images exceptionally well. It also discusses Flux One's limitations, particularly in producing certain art styles like illustrations, where it may not match the finesse of other models like mid-journey. The script includes a demonstration of generating various images using Flux One, comparing the results with those from other AI models, and emphasizing the tool's uncensored nature, which allows for a broader range of creative freedom.

10:03

📈 Flux One's Prompt Adherence and Realism

The script discusses Flux One's performance in prompt adherence, comparing it with mid-journey and Dolly 3 models. It notes that while Flux One does well in incorporating multiple elements from a complex prompt, it may not always match the level of detail and adherence seen with Dolly 3. The paragraph also touches on the model's realism capabilities, suggesting it is on par with mid-journey when using the right prompts. Additionally, it explores the potential of Flux One's uncensored nature, allowing for the generation of copyrighted images and existing IPs, and the possibility of generating not safe for work (NSFW) content in the future.

15:07

🌐 Flux One's Future and Open-Source Potential

The final paragraph of the script looks forward to the future of Flux One, especially considering its open-source nature. It anticipates improvements and customizations by the developer community, which could enhance the model's capabilities. The script also mentions the upcoming text-to-video model based on Flux One, positioning it as a potential open-source alternative to other proprietary tools. The video concludes by acknowledging the contributions of Miguel, also known as Angry Penguin, for his insights into Flux One's strengths and weaknesses, and expresses excitement for the advancements in AI art generation.

Mindmap

Keywords

💡AI Image Generation

AI Image Generation refers to the process where artificial intelligence algorithms are used to create images from scratch based on textual descriptions or other input data. In the context of the video, AI Image Generation is the core theme, as it discusses the capabilities of a new tool called 'Flux.1' developed by Black Forest Labs, which is compared with 'MidJourney' in terms of its ability to generate images that are claimed to be on par or even superior in some aspects.

💡Black Forest Labs

Black Forest Labs is the company behind the development of 'Flux.1', an AI image generation tool. The video script highlights the company's team as having a strong background in creating AI models for image and video generation, including contributions to 'Stable Diffusion' and other models. This establishes the credibility and expertise of the developers in the field of AI image generation.

💡Flux.1 Models

The script mentions three models within the 'Flux.1' suite: 'Flux one Schnell', 'Flux one Dev', and 'Flux one Pro'. Each model offers varying levels of power and cost, with 'Schnell' being the fastest and designed for local development, 'Dev' being more efficient and prompt-adherent, and 'Pro' offering state-of-the-art performance for enterprise solutions. These models are central to the video's exploration of the capabilities and potential applications of 'Flux.1'.

💡Open Source

The term 'Open Source' is used in the script to describe the 'Flux one Schnell' model, which is available under the Apache 2.0 license. This means that the source code is freely available for anyone to use, modify, and distribute. The video emphasizes the implications of this for developers and users, allowing for the creation and sale of tools using 'Flux one Schnell' and the use of generated images for both non-commercial and commercial purposes.

💡Hugging Face

Hugging Face is a platform mentioned in the script where the 'Flux.1' models, specifically 'Schnell' and 'Dev', can be used within 'Hugging Face Spaces'. It serves as an example of a website that has integrated the 'Flux.1' tool, allowing users to access and experiment with AI image generation without the need for local installation or development.

💡Glyph

Glyph is described in the script as an AI workflow builder that allows users to create their own 'Flux' workflows and generate images for free, even using the 'Pro' model. It is highlighted as a platform that simplifies the process of generating images with 'Flux.1', enabling users to input prompts and customize the image generation process through a series of blocks or steps.

💡Prompt Adherence

Prompt Adherence refers to the ability of an AI image generation tool to accurately incorporate all elements mentioned in a textual description into the generated image. The video discusses this concept in the context of comparing 'Flux.1' with 'MidJourney' and 'Dolly 3', noting that while 'Flux.1' performs well, 'Dolly 3' excels in this area by including all prompt elements in the output images.

💡Realism

Realism in the context of AI image generation pertains to the creation of images that closely resemble real-world objects, scenes, or people. The script suggests that 'Flux.1' is designed to produce high-quality, realistic outputs, and it is compared with 'MidJourney' in terms of realism, with the latter still being considered superior in some instances.

💡Uncensored

The term 'Uncensored' is used in the script to describe the 'Flux.1' model's lack of restrictions on the types of images that can be generated, except for a built-in NSFW (Not Safe For Work) filter. This is contrasted with platforms that may impose more stringent content guidelines, and it is suggested that the open-source nature of 'Flux.1' could potentially allow for the removal of these filters in the future.

💡Text to Video Model

The script mentions that 'Flux.1' is not only a text to image model but is also intended to serve as the foundation for an upcoming suite of text to video systems. This indicates that the capabilities of 'Flux.1' will be expanded to include video generation, offering an open-source alternative to proprietary tools like 'Lum's Dream Machine' and 'Runway Gen 3 Sora'.

Highlights

Introduction of Flux.1, a new AI image generating tool developed by Black Forest Labs.

Flux.1 is comparable to Midjourney in capabilities and may outperform it in certain aspects.

The development team behind Flux.1 includes creators of VQ-GAN and Latent Diffusion models.

Three models of Flux.1 with varying power and cost: Schnell, Dev, and Pro.

Flux.1 Schnell is the fastest, open-source model suitable for local development and personal use.

Flux.1 Dev is more efficient and prompt-adherent than Schnell, but for non-commercial use only.

Flux.1 Pro offers state-of-the-art performance designed for enterprise solutions.

Integration of Flux.1 models on various websites for free use.

Using Flux.1 models on Hugging Face and Glyph for generating images.

Flux.1's capability for generating high-quality, realistic images.

Flux.1's strength in handling text within images, making it great for creating logos and memes.

Comparison of Flux.1's prompt adherence with Midjourney and DALL-E 3, noting its middle ground performance.

Flux.1's uncensored nature allows for generating copyrighted images and existing IPs.

Potential of Flux.1 to become an all-encompassing model improving upon Midjourney, DALL-E 3, and Stable Diffusion.

Flux.1's open-source nature enabling community development and customization.

Upcoming text-to-video capabilities of Flux.1, expanding its generative potential.

Recommendation to follow industry experts like Miguel (Angry Penguin PNG) for cutting-edge AI insights.

Tips for improving prompts with Flux.1, emphasizing detailed and descriptive prompts.

Flux.1's current standing as a strong competitor in the AI art generation space.

Anticipation for the future development and community-driven improvements of Flux.1.