Inpainting in Midjourney: A Deep Dive, Secret Commands & Tips for the Long Awaited Feature!

Theoretically Media
22 Aug 202313:58

TLDRThe video provides a comprehensive guide on the newly introduced inpainting feature in Midjourney, a tool for image restoration and enhancement. The host explains how inpainting allows users to restore or modify specific areas of an image without additional input, using contextual clues to determine the changes. The video covers the limitations, secret commands, and fun aspects of using inpainting, demonstrating its power through various examples, such as removing unwanted elements, changing character expressions, and adding props. The host also shares tips on using the feature effectively, including the slider method for complex prompts and the use of specific commands like 'chaos' and 'weird' for creative effects. The summary showcases the potential of inpainting to transform images while maintaining the original context.

Takeaways

  • 🎨 **Inpainting Feature**: Midjourney has introduced a new feature called inpainting, which allows users to restore or modify parts of images without additional input.
  • 🖌️ **Technique Overview**: Inpainting is traditionally used for restoring artwork but is now being used to edit images by highlighting areas and re-prompting within them.
  • ✋ **Extra Fingers Problem**: The feature makes it easier to correct common image generation issues like extra fingers or limbs in generated characters.
  • 🔍 **Very Region Button**: The 'Very region' or inpainting tool is used to select areas of the canvas for Midjourney to reimagine based on contextual clues.
  • 📝 **Remix Mode**: To use inpainting for specific edits, users should enable 'remix mode' through the forward slash settings.
  • 📈 **Image Size Limitation**: Inpainting works best with upscaled images and is most effective when covering a region of 20 to 50% of the image.
  • 🔄 **Undo Function**: There is an 'undo last step' feature available in the inpainting module for users to correct mistakes easily.
  • 🔍 **Free Draw Tool**: The free draw or lasso tool can be used to select areas for inpainting, although it may be finicky and require practice.
  • 🧩 **Combining Tools**: Inpainting can be combined with other tools like zooming out for more editing flexibility, but cannot be used after panning.
  • 🚫 **Limitations**: There are times when inpainting may not work as expected, such as when adding certain props or characters that do not appear in the final image.
  • 🛠️ **Slider Method**: A technique involving repeating parts of the prompt with a colon can sometimes coax the inpainting feature into generating desired elements that were previously ignored.
  • 🔧 **Commands**: Certain commands like 'chaos', 'stylize', 'style', and 'weird' can be used within the inpainting feature to alter the outcome, although 'image waiting' and 'version' do not work.

Q & A

  • What is inpainting in the context of mid-journey features?

    -Inpainting refers to the technique of digitally restoring or editing parts of images or videos, such as removing unwanted elements or fixing damaged areas. In the context of mid-journey, it allows users to highlight an area and re-prompt within that area, enabling the AI to reimagine the scene without additional user input.

  • How does the inpainting feature in mid-journey work?

    -The inpainting feature in mid-journey allows users to select specific areas of their canvas and have the AI reimagine that area based on contextual clues from the surrounding image. Users can access this feature by turning on remix mode and using the 'very region' button to select the area they want to modify.

  • What are the limitations of the inpainting feature?

    -The inpainting feature works best when covering a region of 20 to 50 percent of the image. It is also limited to upscaled images and cannot be used after panning. Additionally, there may be instances where the feature may not produce the desired outcome, such as adding certain props or changing specific elements within the image.

  • How can users ensure they get specific results with the inpainting feature?

    -To get specific results, users can use the 'very region' button to select the area they want to modify and provide a clear and specific prompt. If the AI does not respond as expected, users can try using the slider method by repeating part of the phrase in the prompt or adjusting the prompt to be more explicit.

  • What are some of the secret commands that can be used with the inpainting feature?

    -Some of the secret commands that can be used with the inpainting feature include 'chaos', 'stylize', 'style', and 'weird'. These commands can be added to the prompt to influence the style and outcome of the inpainted area.

  • How can the inpainting feature be used to change the ethnicity or emotions of a subject in an image?

    -By using the inpainting feature and providing a specific prompt, such as 'Korean man' or 'happy', the AI can be directed to change the ethnicity or emotions of a subject in the image. This is done by selecting the area of the subject's face and issuing the new prompt.

  • What is the slider method and how can it be used with the inpainting feature?

    -The slider method is a technique where a colon colon (::) is used to repeat part of the phrase in the prompt. This can help the AI focus on specific elements of the prompt and produce the desired outcome. It can be particularly useful when the AI is not responding to a single prompt as expected.

  • How does the inpainting feature handle lighting and composition?

    -The inpainting feature does a good job of handling lighting and composition, matching the overall lighting and composition within the frame. However, it may not be completely accurate in terms of motivated light, where the light source and its effects are precisely replicated in the inpainted area.

  • What are some potential issues users might encounter when using the inpainting feature?

    -Users might encounter issues such as the feature ignoring their prompts, producing unwanted elements like extra limbs, or not composing the inpainted elements as desired. Additionally, the free draw and lasso tools for selecting areas can be finicky and may require some practice to use effectively.

  • Can the inpainting feature be used with niji images?

    -Yes, the inpainting feature can be used with niji images, allowing for changes to be made to the characters or scenes within the image while maintaining the original background and composition.

  • What is the availability of the PDF mentioned in the transcript?

    -The PDF is available for free on gumroad, but donations are appreciated. Patreon members can access it as part of their membership benefits.

Outlines

00:00

🎨 Introduction to Mid-Journey In-Painting Feature

The video script introduces a new feature called 'in-painting' for the mid-journey stage of image editing. In-painting is a technique used to restore or enhance parts of images or videos. The script explains that this feature allows users to select areas of an image and prompt the AI to reimagine that area based on contextual clues. It also discusses the limitations, such as working best with upscaled images and covering regions of 20 to 50 percent of the image for optimal results. The script provides a step-by-step guide on how to use the in-painting tool, including turning on remix mode and using the free draw and lasso tools for selection. It also mentions that in-painting can correct common issues like extra limbs in generated images.

05:01

🖌️ Advanced In-Painting Techniques and Results

The second paragraph delves into more advanced uses of in-painting, showcasing how it can handle lighting and composition to create a more realistic and contextually accurate image. It demonstrates changing poses, expressions, and adding props to subjects within an image. The script also discusses the challenges of using the lasso tool and the potential for errors, such as the reappearance of unwanted elements like extra limbs. It highlights the power of in-painting to recontextualize a subject while maintaining the background and mentions its compatibility with niji images. The paragraph concludes with examples of changing ethnicities, emotions, and adding elements like ninjas to an action scene.

10:01

🔍 Troubleshooting and Command Usage in In-Painting

The final paragraph addresses some of the challenges and solutions when using in-painting. It talks about instances where in-painting might ignore certain prompts and introduces the 'slider method' as a workaround. This method involves repeating part of the prompt to give the AI a clearer direction. The script also lists commands that can be used within in-painting, such as 'chaos', 'stylize', 'style', and 'weird', while cautioning that 'weird' can produce unexpected results. It provides examples of how these commands can affect the final image, such as adding a clown and an alien to a scene. The paragraph concludes with a reflection on the impressive capabilities of in-painting and an invitation for viewers to share their thoughts and tips.

Mindmap

Keywords

💡Inpainting

Inpainting refers to the technique of digitally restoring damaged or missing parts of images or videos. In the context of the video, it is a feature of the 'mid-journey' software that allows users to highlight an area of an image and re-prompt the system to fill in that area with content that is contextually appropriate. It is used to fix issues like extra limbs or fingers in generated images, and it can also be used creatively to alter scenes, as demonstrated by the host.

💡Mid-journey

Mid-journey is a software or tool that is capable of generating images based on textual prompts. The video discusses a new feature within this tool called 'inpainting', which is used to edit specific areas of an image without affecting the rest. It is a significant update that enhances the software's capabilities for image editing and manipulation.

💡Remix Mode

Remix mode is a setting within the mid-journey tool that, when enabled, allows users to make specific changes to parts of an image using the inpainting feature. The host mentions that to utilize inpainting, one must ensure remix mode is turned on, which suggests it is a crucial step in the process of局部图像编辑 (partial image editing).

💡Very Region

In the context of the video, 'very region' is a term used to describe the button or function within mid-journey that activates the inpainting feature. It is a key component that allows users to select an area of the canvas for the software to reimagine based on surrounding context.

💡Zooming Out

Zooming out is a technique mentioned in the video where the user adjusts the magnification of the image to get a broader view. It is highlighted as a useful method when using inpainting, as it provides more space for the software to generate new content and can lead to more effective results.

💡Free Draw Tool

The free draw tool is a feature within the inpainting module of mid-journey that allows users to manually select areas of the image by drawing a rectangle or using a lasso tool. It is described as being a bit finicky, indicating that it may require some practice to use effectively.

💡Lasso Tool

The lasso tool is a specific selection tool within the inpainting module used to make more complex or irregular selections in an image. The host demonstrates its use by selecting an area around a subject's face to change its expression, showcasing the tool's flexibility in making non-rectangular selections.

💡Chaos and Weird

Chaos and weird are modifiers or commands that can be added to the inpainting prompt to introduce more randomness or unusual elements into the generated content. The host uses these commands to create more dynamic and unpredictable results, such as adding a clown and an alien to a scene.

💡Niji

Niji refers to a style or feature within mid-journey that allows for the generation of images with a particular aesthetic or theme. The host demonstrates the use of inpainting with niji, showing how it can be combined with other features to create unique images.

💡Slider Method

The slider method is a technique mentioned for dealing with the limitations of the inpainting feature when it does not generate the desired content. By repeating parts of the prompt with additional context, the host was able to influence the outcome, eventually achieving the desired result of adding a white wolf to a scene.

💡Style Merging

Style merging is a creative approach where the host combines the styles of two different photographers or artists within mid-journey to generate a unique hybrid style. This technique is used to produce innovative and varied outcomes that blend elements from different sources.

Highlights

Inpainting, a long-awaited feature, has arrived in Midjourney, allowing users to restore or modify parts of images.

Inpainting can correct issues such as extra fingers in images with ease.

The 'Very Region' button is the primary tool for inpainting in Midjourney.

Inpainting uses contextual clues to reimagine selected areas without additional user input.

Remix mode must be enabled for specific inpainting prompts.

Inpainting is most effective with upscaled images.

The tool offers an 'Undo Last Step' feature for easy corrections.

Inpainting works best when covering a region of 20 to 50% of the image.

Combining inpainting with zooming out can yield impressive results.

Inpainting can introduce new elements, such as adding a red balloon to a character's hand.

The tool can change character ethnicities, emotions, and add props.

Inpainting can be finicky with the Lasso tool, requiring practice to master.

The Slider Method can be used to influence inpainting results when the tool isn't responsive.

Commands such as 'chaos', 'stylize', 'style', and 'weird' can be used within inpainting for different effects.

Inpainting maintains the original lighting and composition of the image.

The feature can be used to correct poses and postures in images.

Inpainting is a powerful tool for recontextualizing subjects within an image.

The presenter, Tim, encourages viewers to share their tips and tricks for using inpainting.