BIG Midjourney Update ~ IN-PAINTING Full Breakdown

Future Tech Pilot
22 Aug 202316:27

TLDRThe video provides a detailed walkthrough of the new 'In-Painting' feature in Midjourney, a tool for image editing. The host demonstrates how to use the feature by editing various elements within images, such as changing a man's portrait to show him laughing, adding purple sunglasses to a woman, and even transforming characters into different entities like an android or a king. The importance of using a detailed prompt and the impact of 'chaos' and 'stylize' parameters on the output are highlighted. The video also covers the use of 'remix mode' and the significance of the 'variation' settings. The host shares tips, such as the use of the lasso tool for precise selection and the ability to edit hands and backgrounds without needing a new prompt. The limitations of the feature are acknowledged, including instances where it doesn't work as expected. The video concludes with a demonstration of adding graffiti to a brick wall and the potential for future improvements to the feature.

Takeaways

  • 🎨 **In-Painting Feature**: The new 'In-Painting' feature, also referred to as 'Very by region', allows for localized editing within an image.
  • ✅ **Remix Mode**: To utilize the In-Painting feature effectively, ensure 'Remix mode' is activated by typing 'forward slash remix' in the prompt box.
  • 🖌️ **Selection Tools**: Use the rectangle tool or lasso tool to select the area you wish to edit, with the lasso tool offering more precise control.
  • 📝 **Prompt Adjustments**: You can edit your prompt while the selected region is active, which can drastically change the output.
  • 😆 **Emotion Changes**: The feature can modify facial expressions, such as changing a portrait to show laughing or angry expressions.
  • 👓 **Adding Accessories**: It's possible to add items like purple sunglasses to a subject in the image, although results may vary.
  • 🔍 **High Variation Mode**: For more diverse results, use the 'High Variation' setting, which provides a wider range of outcomes.
  • 👤 **Character Transformation**: The feature can transform characters, such as turning a woman into a man, with impressive results.
  • 🔄 **Undo Function**: If a mistake is made during selection, an 'Undo' button is available to correct it.
  • 📱 **Mobile Compatibility**: The In-Painting feature is accessible via the mobile Discord app.
  • 🚫 **Seeds Irrelevance**: The script suggests that 'seeds' may not impact the In-Painting results, as identical seeds do not produce the same images.

Q & A

  • What is the new feature introduced in the mid-journey update?

    -The new feature introduced is called 'in-painting', which allows users to edit specific regions of an image.

  • How do you activate the remix mode in the software?

    -You can activate the remix mode by typing 'forward slash remix' into the prompt box.

  • What tools can be used to select the area for in-painting?

    -You can use either the rectangle tool or the lasso tool to select the area for in-painting.

  • What is the significance of the 'chaos' parameter in the in-painting feature?

    -A higher chaos value results in more variation in the generated images, providing distinct options to choose from.

  • How does the 'stylize' parameter affect the in-painting results?

    -A lower stylize value makes the look of the generated image more literal and less abstract.

  • What is the purpose of the 'variation options' in the settings?

    -The variation options, which include low and high variation modes, affect the diversity of the results in the in-painting feature.

  • How can you change a character in an image without needing a new prompt?

    -By turning off the remix mode and using the in-painting feature, you can change a character in an image without altering the original prompt.

  • What does the 'seed' parameter do in the in-painting feature?

    -The seed parameter does not seem to affect the results of the in-painting feature as the same seed does not guarantee identical outcomes.

  • How can you use in-painting on an old picture that doesn't have an in-painting button?

    -You can right-click the picture, add a reaction with the envelope emoji, and then use the job ID to bring up the image with the in-painting button.

  • What is the trick to regain access to all the other options after using in-painting?

    -You can use the custom zoom feature, change the zoom level to one, and submit to get fresh pictures that allow access to all options.

  • How effective is the in-painting feature when it comes to adding text or specific details?

    -The in-painting feature can struggle with adding text or very specific details, such as the exact representation of characters.

  • What is the general advice for selecting an area in the in-painting feature?

    -It is advised to select a larger area than expected for better results, especially when trying to add or modify elements within an image.

Outlines

00:00

🎨 Introduction to Mid-Journey's In-Painting Feature

The video introduces a new feature called 'In-Painting by Region' from Mid-Journey, which allows users to edit specific areas of an image. The host demonstrates the feature using a portrait, explaining the importance of enabling 'remix mode' for more control. Various tools like the lasso and rectangle are showcased for selecting areas to edit. The video also covers how to adjust the prompt for different outcomes and the limitations of the feature, such as the disconnect between the smile and eyes in the portrait example.

05:02

🖌️ Exploring In-Painting with Different Prompts and Settings

The host continues to explore the In-Painting feature by changing the eyes of a woman named Marsha to purple and then trying to add purple sunglasses. The segment discusses the importance of being specific in the prompt and the impact of 'low variation' and 'high variation' modes on the output. It also highlights the ability to change characters in an image, such as turning Marsha into a man, and the influence of 'stylize' and 'chaos' parameters on the final image.

10:04

🔄 Advanced Techniques and Mobile App Compatibility

The video moves on to advanced techniques, such as using the panning feature to see more of an image and the custom zoom trick to refresh the image options. It also covers how to use In-Painting to fix specific parts of an image, like hands, without needing to change the prompt. The host shares a tip on selecting larger areas for better results when adding characters to a scene. The segment ends with a note on the feature's compatibility with the mobile Discord app and the unpredictability of the results.

15:05

🚧 Testing In-Painting with Text and Shadows

The final paragraph involves testing the In-Painting feature with text and shadows. The host attempts to add graffiti text to a brick wall and, while the text doesn't appear, the spray paint effect looks impressive. There's also an experiment with casting shadows using the feature, which shows an understanding of light sources and object obstruction. The video concludes with the host's willingness to explore more and an invitation for viewers to share their findings in the comments.

Mindmap

Keywords

💡Mid-journey feature

The term 'mid-journey feature' refers to a new capability or update that has been introduced to an existing software or application during its lifecycle. In the context of the video, it denotes the recent addition of an 'in-painting' feature to an AI image generation tool, which allows users to modify or add elements to existing images in a more controlled manner.

💡In-painting

In-painting is a technique used in image processing that involves filling in missing or selected parts of an image. The video discusses the new 'in-painting' feature, which the presenter refers to as 'very by region', a term that seems to be a play on words or a specific naming by the developers. It allows users to edit specific areas of an image by adding or changing elements within it.

💡Remix mode

Remix mode is a setting within the AI image generation tool that, when enabled, allows users to modify the original prompt used to generate an image. In the video, the presenter suggests turning on remix mode by typing 'forward slash remix' into the prompt box, which then enables more dynamic and varied edits to the image through the in-painting feature.

💡Lasso tool

The lasso tool is a selection tool in image editing software that allows users to make freehand selections of areas in an image. The video demonstrates the use of the lasso tool to select specific parts of an image for in-painting, providing more precise control over the areas to be edited.

💡Variation options

Variation options refer to the different stylistic outcomes that can be generated from a single prompt or image edit. The video mentions 'low variation mode' and 'high variation mode', which affect the results of the in-painting feature, offering users a choice between more uniform or more diverse image outcomes.

💡Chaos parameter

The chaos parameter is a setting that influences the level of randomness or variation in the generated images. A higher chaos value results in more varied and unpredictable outcomes, as demonstrated by the presenter when they adjust the parameter to achieve different levels of detail and variation in the in-painted images.

💡Stylize parameter

The stylize parameter is another setting that affects the style and appearance of the generated images. Lowering the stylize value results in more literal interpretations of the prompt, while higher values allow for more abstract and artistic representations, as shown in the video when the presenter adjusts this parameter.

💡Seed

In the context of AI image generation, a seed is a specific input or setting that determines the initial conditions for the generation process. The video explores the concept of using seeds to try and achieve consistent results, but the presenter finds that the seed does not seem to affect the outcome of the in-painting feature.

💡Job ID

A job ID is a unique identifier assigned to a particular task or operation within a software system. In the video, the presenter describes how to use a job ID to retrieve and manipulate old images with the new in-painting feature, demonstrating a method to apply the feature to images that do not initially display the in-painting option.

💡Panning feature

The panning feature allows users to expand the viewable area of an image, often to include more of the surrounding context. The video shows how the panning feature can be used in conjunction with in-painting to edit areas of an image that were previously outside the frame.

💡Custom zoom

Custom zoom is a function that enables users to adjust the zoom level of an image to focus on specific details or to change the field of view. The presenter in the video uses the custom zoom feature to regenerate an image at a different zoom level, which then allows for additional editing options, including in-painting.

Highlights

Midjourney's new in-painting feature has been released, allowing for more detailed image editing.

The feature is called 'very by region', which is similar to in-painting.

Tips and tricks are provided for using the in-painting feature effectively.

Remix mode should be turned on for the best results with in-painting.

The lasso tool provides better control over the area to be edited.

In-painting allows for the selection and editing of specific areas within an image.

The feature can change facial expressions, such as adding a laughing expression to a portrait.

In-painting is not perfect and results can vary, sometimes not meeting expectations.

Writing a longer, more specific prompt can improve the in-painting results.

High variation mode can produce more diverse results for selection.

Changing characters, such as turning a woman into a man, is possible with in-painting.

The stylize and chaos parameters can significantly affect the outcome of in-painted images.

In-painting can introduce new characters into an existing scene without the need for a new prompt.

The feature works well for editing hands and other specific parts of an image.

In-painting can be used on old pictures by copying the job ID and using the 'show' command.

Seeds do not seem to affect the outcome of in-painting, making them less relevant.

In-painting can be used in combination with other Midjourney features, such as panning.

The in-painting feature is still in its early stages and subject to change and improvement.

The feature allows for creative experimentation and can produce impressive results with the right prompts and settings.