Flux Pro 1.1 is Here! Let's test it.

All Your Tech AI
3 Oct 202411:32

TLDRFlux Pro 1.1, developed by Black Force Labs, has been released, promising six times faster image generation and improved quality over its predecessor. The update, previously tested as 'blueberry', outperforms other models in quality and aesthetics. It's already integrated into Pixel Dojo, allowing users to create images with various Flux-powered options. The video tests Flux Pro 1.1 with old prompts, comparing the results to Flux Pro 1.0, showing distinct differences and improved adherence to prompts. Flux Pro 1.1 also supports up to 2K image resolution, set to be a game-changer in the AI image generation space.

Takeaways

  • 🚀 Flux Pro 1.1 has been released by Black Force Labs, offering significant improvements over its predecessor.
  • 🔍 The new version, code-named 'Blueberry', has been tested and shown to outperform other models in quality and aesthetics.
  • ⏱️ Flux Pro 1.1 is six times faster in image generation compared to the older model.
  • 🖼️ It enhances image quality, prompt adherence, and diversity, with particular improvements in character and human face rendering.
  • 📊 Flux Pro 1.1 outperformed other models like IDE diagram, Mid Journey 6.1, Flux Dev, and Dolly 3 in tests.
  • 🆕 An upcoming feature of Flux Pro 1.1 is the capability to generate images up to 2K resolution.
  • 🎨 Pixel Dojo has integrated Flux Pro 1.1, allowing users to create images using various Flux-powered options.
  • 📸 Users can compare the visual differences between Flux Pro 1.0 and 1.1 by reusing old prompts to see improvements.
  • 🕷️ Flux Pro 1.1 shows variations in prompt adherence, such as generating a less angry-looking tarantula when the 'extremely angry' part of the prompt was removed.
  • 🌐 The API for Flux Pro 1.1 currently does not support adjustments to the guidance scale or steps, which are locked to default values.
  • 📝 There have been significant changes to the large language model and text embeddings that transform prompts into images, affecting how the model is steered.

Q & A

  • What is the main announcement made by Black Force Labs?

    -Black Force Labs announced the release of Flux 1.1 Pro and their API, which is a significant update to their flux-powered tools.

  • How does Flux 1.1 Pro improve upon the previous model?

    -Flux 1.1 Pro offers six times faster generation, improved image quality, better prompt adherence, and increased diversity according to the announcement.

  • What was the code name given to the model that was being tested by Black Force Labs prior to the announcement?

    -The model being tested under the code name 'blueberry' turned out to be Flux 1.1 Pro, as revealed by the announcement.

  • What improvements does Flux 1.1 Pro bring to character and human face depiction?

    -Flux 1.1 Pro improves the depiction of characters and human faces, hopefully eliminating the 'flux chin' issue seen in older models.

  • What new features are mentioned as coming soon for Flux 1.1 Pro?

    -The announcement mentioned that Flux 1.1 Pro will soon support up to 2K image resolution.

  • Where can Flux 1.1 Pro be tested immediately after its announcement?

    -Flux 1.1 Pro can be tested immediately on Pixel Dojo, where it has been integrated into the Flux Image Creator tool.

  • What is the significance of the 'luras' support in the new Flux model?

    -The new Flux model supports up to 20 different custom luras models directly within the flux generations, enhancing customization capabilities.

  • How does the performance of Flux 1.1 Pro compare to other models like IDE, mid Journey, and Dolly?

    -Flux 1.1 Pro outperforms other models such as IDE, mid Journey, and Dolly, as demonstrated by the impressive results in the announcement.

  • What is the user experience like when testing Flux 1.1 Pro on Pixel Dojo?

    -Users can create images using prompts, select aspect ratios, and choose from various models including Flux Pro 1.1, and compare the results with previous models.

  • What changes have been made to the large language model in Flux 1.1 Pro?

    -The large language model in Flux 1.1 Pro has been updated, which affects how prompts are transformed into images, requiring users to adapt their prompting strategies.

Outlines

00:00

🚀 Introduction to Flux 1.1 Pro and Black Force Labs API

The video script begins with the excitement surrounding the release of Flux 1.1 Pro and the Black Force Labs API by the company behind Flux. The script mentions that a mysterious model code-named 'blueberry' has been outperforming others in quality and aesthetics, and it is now confirmed to be Black Forest Labs testing Flux 1.1 Pro. The announcement claims improvements in speed and quality, with Flux 1.1 Pro being six times faster than its predecessor and enhancing image quality, prompt adherence, and diversity. There's also a mention of an issue known as 'flux chin' and hopes that the new model has resolved it. The script highlights impressive results from tests under the 'blueberry' code name, outperforming other models like IDE diagram, mid Journey 6.1, flux Dev, and Dolly 3. The script also anticipates the upcoming feature of 1.1 Pro supporting up to 2K image resolution. The narrator has integrated Flux 1.1 Pro into Pixel Dojo, allowing users to create images using various Flux-powered options, including the new Flux Image Creator which supports different models like Flux Schell, Laura, Hyper 8, Hyper 16, and Flux Pro 1.0. The script ends with a plan to test Flux Pro 1.1 using old Flux Pro prompts to see if there's a visual difference.

05:03

🖼️ Comparing Flux Pro 1.0 and 1.1 Pro Image Generations

The second paragraph of the script discusses the results of generating images using Flux Pro 1.1 Pro compared to Flux Pro 1.0. The narrator tests the new model by using prompts that were successful with the older model and compares the visual differences. The script describes the process of generating images through Pixel Dojo's community gallery, where the narrator has already generated some 'mind-blowing' images with Flux Pro. The narrator then selects several images from the gallery, copies their prompts and seeds, and再生们 using Flux Pro 1.1 Pro to compare the results. The comparisons show that while some images retain a similar aesthetic, others have a noticeably different look, with one being described as more like a PlayStation 5 game or cartoony CGI. The narrator also notes the lack of support for steps or guidance in the new model, which is locked in the API. The paragraph concludes with the narrator's observations on the differences in prompt adherence and image quality between the two versions of Flux Pro.

10:03

📝 Exploring Prompt Adherence and Text Generation with Flux Pro 1.1

The final paragraph of the script delves into the differences in prompt adherence and text generation capabilities of Flux Pro 1.1 compared to its predecessor. The narrator notes that there have been significant changes to the large language model that transforms prompts into images, which will require time to understand and adjust to. The script includes an example of creating an image of a Coca-Cola bottle with spider features, which turned out well, and compares it to an original image generated with Hyper flux 16. The narrator invites viewers to check out Flux 1.1 Pro on Pixel Dojo and share their thoughts in the comments. The video ends with the narrator, Brian, signing off until the next time.

Mindmap

Keywords

💡Flux Pro 1.1

Flux Pro 1.1 is an advanced generative AI model designed to create images based on text prompts. In this video, it is described as an upgraded version of the original Flux Pro, offering six times faster generation speeds and improved image quality. The release of Flux Pro 1.1 is a significant development in the field of AI-powered image generation, especially with its enhanced ability to generate more realistic and varied images compared to its predecessors.

💡Pixel Dojo

Pixel Dojo is a platform mentioned in the video where users can create and explore AI-generated images. The video showcases how Flux Pro 1.1 has been integrated into Pixel Dojo, allowing users to access and experiment with different AI models, including Flux Pro 1.1, for image creation. It's a hub for AI-driven art and creative exploration.

💡API

API (Application Programming Interface) refers to the set of tools and protocols that allow different software systems to communicate. In the video, Black Force Labs has released the Flux 1.1 Pro API, enabling developers and users to integrate the new Flux model into their own applications or systems, such as Pixel Dojo. This integration is key for users who want to automate or customize their image generation workflows.

💡Image Quality

Image quality refers to the clarity, sharpness, and realism of an image. Flux Pro 1.1 is highlighted for its improved image quality compared to earlier models, particularly in terms of character faces and detail accuracy. The video shows comparisons between images generated by Flux Pro 1.0 and 1.1, demonstrating how the new version produces more lifelike and visually appealing results.

💡Prompt Adherence

Prompt adherence describes how well the generated image matches the user's original text prompt. The video mentions that Flux Pro 1.1 improves prompt adherence, ensuring that the generated images align more closely with the input descriptions. For instance, when creating a digital artwork of a woman with specific attributes, Flux Pro 1.1 returns images that are more faithful to the provided prompt than previous versions.

💡Model Comparison

Model comparison in the video refers to the process of evaluating different AI models based on their performance and output quality. The script compares Flux 1.1 Pro to other models like Mid Journey, Dolly 3, and Stable Diffusion, showing how Flux 1.1 outperforms them in both speed and quality. This comparison helps illustrate the advancements made in Flux Pro 1.1.

💡Lora Support

Lora support refers to the ability to integrate different custom models (known as Loras) into the image generation process. In the video, Flux Pro 1.1 on Pixel Dojo allows users to select from various Loras, offering more flexibility and customization in the generated results. This feature enhances the creativity and uniqueness of the images produced by Flux Pro.

💡Resolution

Resolution refers to the detail an image holds, usually measured in pixels. The video mentions that Flux Pro 1.1 will soon support up to 2K image resolution, which is a significant upgrade over previous models. Higher resolution results in sharper and more detailed images, which is especially important for users who require high-quality outputs for professional or artistic purposes.

💡Text-to-Image Generation

Text-to-image generation is the process of using natural language descriptions (prompts) to generate corresponding visual content. The video focuses on how Flux Pro 1.1 excels in text-to-image generation, producing creative and realistic images from detailed prompts. This technology is a central theme in the video, as the user demonstrates how different prompts result in varied and impressive image outputs.

💡Custom Seeds

Custom seeds refer to specific numeric values that influence the randomization process during image generation. By using the same seed, users can recreate identical or very similar images. In the video, the narrator tests Flux Pro 1.1 by using the same seed and prompt from earlier Flux Pro 1.0 images, allowing for a controlled comparison of the outputs. This demonstrates the consistency and variability that can be achieved with the same input parameters.

Highlights

Flux Pro 1.1 has been released, offering significant improvements over previous versions.

Black Force Labs announced Flux 1.1 Pro and the Black Force Labs API, which is a major development in the AI image generation field.

The new Flux Pro 1.1 is codenamed 'blueberry', which was previously a mystery model outperforming others in quality and aesthetics.

Flux 1.1 Pro provides six times faster generation compared to the older model.

Image quality, prompt adherence, and diversity have been improved in Flux 1.1 Pro.

Flux 1.1 Pro promises better quality in character and human faces, potentially eliminating the 'flux chin' issue.

Tests under the code name 'blueberry' showed impressive results, outperforming IDE diagram, Mid Journey 6.1, Flux Dev, and other models.

Dolly 3 and Stable Diffusion 3 medium were significantly outperformed by Flux Pro 1.1.

Flux Pro 1.1 will soon support up to 2K image resolution, a significant upgrade.

Pixel Dojo has integrated Flux Pro 1.1, making it available for users to create images.

Flux Image Creator allows users to create images using various Flux-powered models, including the new 1.1 version.

Flux Pro 1.1 was tested with old Flux Pro prompts to compare visual differences.

Pixel Dojo's community gallery contains thousands of images generated with Flux Pro, showcasing the quality of the model.

Flux Pro 1.1 generates images that are quite different from the original Flux Pro, with a more cell-shaded and CGI-like appearance.

Flux Pro 1.1 does not currently support steps or guidance adjustments in the API.

Comparisons between Flux Pro 1.0 and 1.1 show differences in prompt adherence and image aesthetics.

Documentation indicates significant changes to the large language model in Flux Pro 1.1, affecting how prompts are transformed into images.

Flux Pro 1.1 is available on Pixel Dojo, offering over 20 generative AI tools in one subscription.