V5, V6, V7? | Midjourney Office Hours Recap Feb 8th 2023 | Midjourney News

Future Tech Pilot
9 Feb 202303:43

TLDRIn the mid-journey office hours recap, David discusses the introduction of wider aspect ratios and acknowledges that changes in algorithms can lead to variations in output quality. He emphasizes the iterative nature of improvements, with versions 5 through 7 in consideration, highlighting the potential for higher resolution and clearer images. David also addresses the challenges of balancing cost and accessibility for new upscalers due to their high GPU usage. Additionally, he mentions the team's reluctance to offer a digital magazine, valuing the physical aspect, and shares updates on website improvements, including a fast search feature.

Takeaways

  • 📅 David discussed updates from the mid-journey office hours on February 8th.
  • 🌟 The wider aspect ratios feature was released, allowing for more flexibility in image proportions.
  • 🔄 Service shifts and algorithm changes may lead to improvements overall but can also negatively affect some edge cases.
  • 🚀 Small updates and bug fixes are expected for the niji Journey platform.
  • 🧠 Training the same model twice can yield different results, highlighting the complexity and variability of AI models.
  • 🎨 The team is considering moving forward with version 5 due to dissatisfaction with the current tiling version and its sharpness.
  • 👁️ Version 5 promises higher resolution and clearer images without the sharpness bug present in version 4.
  • 🔜 Version 5 could be released soon, but its release timeline depends on whether any unforeseen problems arise during testing.
  • 💡 The team is also exploring the idea of incorporating different upscalers into various pricing plans to manage costs and accessibility.
  • 📚 International shipping costs are too high to offer the magazine for free, so a plan to offer the first issue complimentary with subsequent shipping fees is being considered.
  • 🔍 A fast search feature, along with other improvements like in-painting and a zoom-out feature, are in development for the website.

Q & A

  • What new feature was mentioned at the beginning of the office hours recap?

    -The new feature mentioned at the beginning is the wider aspect ratios that were released, allowing users to utilize any aspect ratio they can think of.

  • What is a potential downside of changing the service's algorithm?

    -A potential downside is that while the overall service may improve, some edge cases might get worse due to the changes in the algorithm.

  • What does David say about training the same model twice?

    -David mentions that training the same model twice will result in different outcomes, highlighting the difficulty in controlling for variations.

  • What advice does David give regarding attachment to specific styles or outputs?

    -David advises not to get attached to anything specific, such as a style or output, as it may not always be available as they continue to make changes and improvements.

  • What is the team's current concern with the version of tiling they have?

    -The team is not satisfied with their current version of tiling and are considering putting more effort into version 5, which shows promise in providing higher resolution and clearer images.

  • Why does the version 4 of the service seem less sharp?

    -Version 4 is not as sharp as it should be due to an unknown bug, which the team is trying to address in the upcoming version 5.

  • What is the team's strategy for releasing new versions?

    -The team aims for consistent releases that can improve more quickly, rather than trying to include all possible features in each version.

  • How might the company handle resolution differently in the future?

    -The company might handle resolution differently by incorporating different upscalers into different price plans, which could relate to raw pixel count rather than image manipulation.

  • What is the concern regarding the cost of new upscalers?

    -The concern is that the new upscalers will be expensive to run, costing a lot of GPU minutes, and the company is considering how to make this technology accessible without users in lower-tier plans quickly depleting their minutes.

  • What is the update on the mid-journey magazine?

    -The update is that international shipping costs are too high to offer the magazine for free, so they are considering making the first issue free and then requiring users to cover shipping costs for subsequent issues.

  • What new features are being worked on for the website?

    -The website is in the process of adding a very fast search feature, improving in-painting and zoom-out features, and they are actively hiring more people to support these developments.

Outlines

00:00

🚀 Updates and Future Plans in Mid-Journey Office Hours

The video script discusses the recent updates and future plans shared during the Mid-Journey Office Hours on February 8th. David highlights the introduction of wider aspect ratios, which allows for more flexibility but may also lead to some edge cases getting worse due to algorithm changes. He emphasizes the iterative nature of improvements and advises against becoming too attached to specific styles or outputs. The team is considering a version of tiling but is not satisfied with the current results, potentially focusing on version 5 instead. Early models for higher resolution and clearer images show promise, and the team is already looking at versions 5, 6, and 7 for consistent and rapid improvements. However, the release timeline for version 5 remains uncertain, depending on whether any unforeseen problems arise. David also talks about potential changes in handling resolution in the future, such as incorporating different upscalers into various price plans. The team is aware of the high cost of new upscalers and is considering ways to make them accessible. Lastly, the script mentions the challenges with international shipping for the Mid-Journey magazine and the upcoming features for the website, including a fast search function, inpainting, and a zoom-out feature.

Mindmap

Keywords

💡aspect ratios

Aspect ratios refer to the proportional relationship between the width and height of an image or video frame. In the context of the video, it means that users now have the flexibility to use various aspect ratios for their content, allowing for more creative freedom and diverse visual presentations. The mention of wider aspect ratios being released last week indicates an update that broadens the service's capabilities.

💡algorithm

An algorithm is a set of rules or instructions for solving problems or accomplishing tasks, often used in computing and data processing. In the video, the term is used to describe the underlying mechanisms that determine how the service functions, especially when it comes to adjusting to new aspect ratios. Changes in the algorithm can lead to improvements but may also affect certain edge cases negatively, highlighting the complexity of software development and maintenance.

💡Niji Journey

Niji Journey appears to be the name of a product or service being discussed in the video. It is mentioned in the context of needing bug fixes and updates, suggesting that it is a software application or platform that is currently in use and being actively developed and improved upon. The speaker's comments about training models and versioning indicate that Niji Journey involves machine learning or AI components.

💡versioning

Versioning is the process of keeping track of changes to a product or system over time, typically through the assignment of version numbers or labels. In the video, the speaker discusses training the same model twice and getting different results, which underscores the importance of version control in software development. This ensures that each iteration of Niji Journey is documented and that improvements or changes can be traced and managed effectively.

💡tiling

Tiling, in the context of the video, likely refers to a method of arranging images or content to fill a space without gaps or overlaps, much like tiles on a floor or walls. The speaker mentions that the team has a version of tiling but is not satisfied with it, indicating an ongoing effort to refine and enhance the functionality of their service. This reflects a commitment to quality and user experience in the development process.

💡resolution

Resolution typically refers to the clarity or sharpness of an image, determined by the number of pixels or the level of detail it contains. In the video, the discussion about higher resolution and clearer images suggests that the team is working on improving the visual quality of their outputs. The mention of different upscalers for different price plans indicates a potential strategy for offering varying levels of detail or clarity based on user needs and payment tiers.

💡GPU

GPU stands for Graphics Processing Unit, a specialized electronic circuit designed to rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display device. In the context of the video, GPUs are crucial for running the upscalers, which are processes that increase the resolution of images. The speaker expresses concern about the cost of these upscalers in terms of GPU minutes, highlighting the computational expense and the need to balance accessibility with the cost of providing high-quality services.

💡rating system

A rating system is a method of evaluating and quantifying the quality or performance of something, often used to provide feedback or make comparisons. In the video, the speaker mentions that a new rating system will be unveiled, indicating an upcoming feature or update that will allow users or the service itself to assess and rank content. This could be a way to improve the curation and recommendation of images or to gather user feedback for further development.

💡mid-journey magazine

The term 'mid-journey magazine' refers to a publication that is presumably related to the company or service discussed in the video. The speaker talks about the challenges of international shipping costs and the consideration of making the first issue free, with subsequent issues requiring payment for shipping. This suggests that the magazine is a physical product that the company is producing as part of its offerings, valuing the tangible aspect of content consumption.

💡website

In the context of the video, the website is the digital platform where the company provides information and services to its users. The speaker mentions that they are still working on improving the website, with plans for a fast search feature and other enhancements like in-painting and a zoom-out feature. This indicates an ongoing effort to enhance user experience and provide more functionality on their online platform.

💡upscalers

Upscalers, in the context of the video, refer to software or algorithms that increase the resolution or quality of digital images. The speaker discusses the cost and potential impact of new upscalers on their service, indicating that these tools are significant in improving the output quality. The mention of expense and GPU minutes suggests that these upscalers are advanced and resource-intensive, requiring careful consideration for implementation and pricing.

Highlights

Introduction of wider aspect ratios in the latest release, allowing for more creative flexibility.

Acknowledgment that service may shift and some edge cases could worsen with new aspect ratio implementation.

Upcoming small updates to improve the existing system, including bug fixes.

Fascinating insight that training the same model twice can yield different outcomes, indicating the complexity of the AI system.

Advice not to get too attached to specific styles or outputs as the platform continues to evolve and change.

Discussion about the potential of a new version of tiling, with the current version not meeting expectations.

Exciting prospects of higher resolution and clearer images in the promising early models.

Admission that version 4 lacks sharpness, and the team is looking ahead to versions 5 and beyond.

Version 5 is showing no signs of the bug present in version 4, offering more detailed and accurate outputs.

The team is considering consistent, rapid releases to improve the system, rather than overloading each version with features.

Initial images for version 5 look promising, but the team is cautious about exploring all edge cases before release.

Version 5 could be released soon, but the timeline depends on whether any unforeseen problems arise during further testing.

Plans to introduce a new rating system, but no images will be shared ahead of the official release.

David warns that despite technological advancements, there are no guarantees in release windows due to the unprecedented nature of the AI.

Consideration of different upscaling methods in future price plans, potentially offering different levels of pixel enhancement.

Concerns about the affordability and feasibility of new upscalers due to high GPU costs.

The company's focus on making the technology accessible, despite the high operational costs.

Challenges with international shipping costs for the physical magazine, with plans to offer the first issue for free and charge for subsequent shipping.

Preference for a physical magazine over a digital one, emphasizing the importance of tangible objects in sharing and community building.

Ongoing website improvements, including a fast search feature, inpainting, and a zoom-out function.

Efforts to hire more staff to support the development and maintenance of the platform.