『Reference Only』で“同じ顔”や”同じキャラクター”をキープできる。凄すぎるControlNetの新機能

AI FREAK - 最新のAIツールをご紹介
16 May 202309:17

TLDRControlnet introduces a groundbreaking new feature called 'Reference Only,' which allows users to adjust various elements while maintaining the same person or character in an image. This technology is particularly impressive as it enables the creation of cartoons, animations, and frame-by-frame GIF movies with ease. The process is simple: upload an image, enable the feature, and select 'Reference Only' from the Preprocessor. The result is a visually similar image that can be further refined by adjusting the batch count and control modes. The feature is highly versatile, compatible with different models, and can be used to create high-quality animations and detailed illustrations. By combining 'Reference Only' with other functions like Multi ControlNet, users can achieve even more precise and flexible adjustments, making it easier than ever to generate the desired image. This feature significantly lowers the barrier to creating an ideal image and opens up new possibilities for creative expression.

Takeaways

  • 🆕 ControlNet has introduced a new feature called 'Reference Only' that allows users to adjust different elements while keeping the same face or character.
  • 📸 To use 'Reference Only', you need to enable it and select it from the Preprocessor section in the ControlNet interface.
  • 🐾 An example given in the script is generating images of a dog where the face remains consistent across different prompts and variations.
  • 🎨 Users can adjust the 'Control Mode' to prioritize their prompt, which can result in more consistent character generation.
  • 🏃 By adding specific actions like 'Running' to the prompt, the model can generate images with the same face but different poses.
  • 🎭 The 'Reference Only' feature is versatile and can be used to create cartoons, animations, and even frame-by-frame GIF movies.
  • 📈 Increasing the batch count can improve the consistency and accuracy of the generated images.
  • 👩‍🎤 The technology is compatible with illustration systems, facilitating more detailed depictions in animations and cartoons.
  • 👱‍♀️ It's possible to change specific attributes like hair color while keeping the character's face consistent.
  • 🍎 However, the accuracy of reflecting additional objects or actions in the image may require more specific prompts.
  • 🎥 For live-action images, 'Reference Only' can generate faces that are very similar to the original, offering flexibility in pose and clothing changes.
  • 👗 The feature can be used to create a photo collection of a particular model by keeping the face consistent while changing clothing styles.
  • 🔍 Multi ControlNet is a feature that allows for the application of multiple controlnets to find suitable images for specific tasks, like holding an object.
  • 🔧 Adjusting 'Control Weight' and Modes can help refine the generated image to better match the original face and desired outcome.
  • 📈 The 'Reference Only' feature significantly lowers the barrier for generating ideal images and opens up many creative possibilities.

Q & A

  • What is the new feature 'Reference Only' in ControlNet?

    -The 'Reference Only' feature in ControlNet allows users to adjust different elements in an image while keeping the same person or character intact. It is a part of the ControlNet's advanced technologies that enable the generation of images with consistent character appearances.

  • How does the 'Reference Only' feature work?

    -To use 'Reference Only', you first upload an image into the ControlNet section, check 'Enable' and 'Pixel Perfect', select 'Reference Only' from the Preprocessor options, and then press 'Generate'. The feature ensures that the generated image retains the same face or character as the original.

  • What happens if 'Reference Only' is not visible in the ControlNet interface?

    -If 'Reference Only' is not visible, it indicates that ControlNet needs to be updated to include this new feature.

  • Can the 'Reference Only' feature be used to create variations in the generated images?

    -Yes, the feature allows for variations in mood, background, and other elements while maintaining the same character face. Users can also adjust different modes in Control Mode to prioritize the prompt and achieve a more similar character image.

  • How can the 'Reference Only' feature be used to create animations or cartoons?

    -By increasing the batch count, users can create frame-by-frame GIF movies. The feature is versatile enough to be used in creating cartoons and animations, allowing for the generation of a sequence of images with the same character in different poses or actions.

  • What is the XYZ Plot section in the Scripts of ControlNet?

    -The XYZ Plot section is a feature in ControlNet that, while not detailed in the script, is implied to be useful for generating variations of images based on different models while keeping the character consistent.

  • How does the 'Reference Only' feature perform with illustrative images?

    -The feature is quite compatible with illustration systems, facilitating detailed depictions and allowing for the creation of cartoons and animations with ease.

  • Can the 'Reference Only' feature be used to change the color of a character's hair while keeping the character's face?

    -Yes, the feature allows for changes in hair color and other attributes while maintaining the character's face similarity. It may require increasing the batch count and careful selection to improve accuracy.

  • What happens if the original image is not in the right pose or size for certain features like holding an apple?

    -The feature may not work as expected, and the accuracy might not be as good. In such cases, adjusting the prompts, size, and trying again with a more suitable image can help achieve the desired result.

  • How can Multi ControlNet be utilized to improve the accuracy of generated images?

    -Multi ControlNet allows for the application of multiple controlnets, searching for a suitable image to match the desired outcome. It can be used to find a better pose or element, like holding an apple, and then combine it with the original image to improve the result.

  • How can the 'Control Weight' be adjusted to correct the atmosphere of the generated image?

    -By lowering the Control Weight on the Openpose side and changing the Mode, the face can be made to resemble the original image more closely. This adjustment helps in retaining the original face while changing other aspects like the pose.

  • What are the potential applications of the 'Reference Only' feature?

    -The 'Reference Only' feature can be used to generate a wide range of images, from creating photo collections of specific models to facilitating the production of detailed illustrations and animations. It lowers the barrier for generating ideal images and opens up many creative possibilities.

Outlines

00:00

🐕 Introduction to Controlnet's Reference Only Feature

The first paragraph introduces a new feature in controlnet called 'Reference Only'. It allows users to adjust different elements of an image while maintaining the same person or character. The feature is considered one of the most impressive in controlnet and is encouraged for trial. A step-by-step guide on how to use the feature is provided, starting with uploading an image of a dog and adjusting settings in the controlnet section. The process involves enabling options, selecting 'Reference Only', and generating the image. The resulting images are compared to showcase the similarity and variations. The paragraph also discusses changing modes and prompts to achieve desired results, such as adding a 'running' prompt to the image. The versatility of controlnet is highlighted, mentioning its potential for creating cartoons, animations, and GIF movies. The feature's compatibility with different models is tested by applying the same prompts to generate images of dogs. Lastly, the feature's effectiveness with illustrative images is tested by adjusting hair color and adding objects like an apple to the image.

05:06

🎨 Exploring Multi ControlNet and Image Adjustments

The second paragraph delves into the use of Multi ControlNet, which allows for the application of multiple controlnets to refine image generation. It begins by demonstrating the high-quality and flexibility of the generated images, particularly in animation. The paragraph then explores the use of inpaint to change poses and clothing while maintaining the same face. The challenge of accurately reflecting the intent of holding an apple in an image is addressed, and adjustments to prompts and image size are suggested to improve results. The paragraph also discusses the use of Openpose settings to adjust poses and the importance of adjusting Control Weight to retain the original face's elements. The process of generating an image with a specific pose and applying changes to the face and prompts is detailed. The paragraph concludes by emphasizing the lowered barrier for generating ideal images through the combination of 'Reference Only' with other functions, and encourages viewers to explore further applications of the technology. The video ends with a call to action for viewers to subscribe to the channel and engage with the content.

Mindmap

Keywords

💡Reference Only

Reference Only is a new feature in ControlNet that allows users to adjust different elements of an image while maintaining the same person or character's appearance. It is considered one of the most remarkable technologies within the platform. In the video, it is used to generate images of a dog and a woman, keeping their faces consistent across various prompts and modifications, showcasing its ability to maintain a character's identity while altering other aspects of the image.

💡ControlNet

ControlNet is a technology or platform that offers various features for image generation and manipulation. The script highlights its versatility and the introduction of the Reference Only feature, which is a significant addition to its capabilities. The video demonstrates how ControlNet can be used to create consistent character images across different prompts and settings, emphasizing its utility in generating a series of related images.

💡Pixel Perfect

Pixel Perfect is a term often used to describe images that are sharp and clear down to the smallest detail. In the context of the video, checking the Pixel Perfect box ensures that the image manipulations performed by ControlNet are of high quality and maintain the fidelity of the original image. It is mentioned as part of the process of using the Reference Only feature to generate images.

💡Preprocessor

A Preprocessor in the context of the video is a part of the ControlNet system that allows users to select specific features or modes, such as Reference Only, before generating an image. It is an essential step in the process of image generation, as it determines how the input data will be handled and what specific features will be applied to the output image.

💡Batch Count

Batch Count refers to the number of images generated in a single operation. In the video, increasing the batch count allows for the creation of multiple images based on the same input, which helps in observing variations and ensuring consistency in the character's face across different images. It is used to demonstrate the consistency of the Reference Only feature.

💡Control Mode

Control Mode is a setting within ControlNet that allows users to adjust the level of control they have over the image generation process. The video mentions changing Control Mode to 'My prompt is more important,' which increases the priority of the user's input, resulting in a more accurate representation of the desired image.

💡Multi ControlNet

Multi ControlNet is a feature that enables the application of multiple ControlNet processes to a single image. It is used in the video to combine different aspects of image generation, such as pose and facial features, to create a more complex and detailed output. It is mentioned as a way to improve the accuracy and flexibility of the generated images.

💡Openpose

Openpose is a technology used within the ControlNet platform to apply or adjust the pose of a character in an image. In the video, it is used in conjunction with the Reference Only feature to change the pose of a character while retaining their facial features, demonstrating the ability to manipulate specific aspects of an image independently.

💡Control Weight

Control Weight is a parameter within ControlNet that allows users to adjust the influence of certain features or settings on the image generation process. In the video, adjusting the Control Weight helps to fine-tune the image, ensuring that the original face is retained while the desired changes, such as pose or background, are applied.

💡Inpaint

Inpaint is a term used to describe a process of editing an image to fill in or alter certain areas. In the context of the video, it is mentioned as a method to change the pose or clothes of a character in an image. However, the video suggests that using Reference Only in ControlNet provides more flexibility and control over these adjustments.

💡Illustrative Images

Illustrative Images refer to images that are created in a style typically associated with illustrations, such as cartoons or animations. The video discusses the compatibility of the Reference Only feature with illustrative styles, showing how it can be used to generate a series of images with consistent character designs suitable for creating cartoons and animations.

Highlights

Introducing a new feature called 'Reference Only' to ControlNet that allows adjusting different elements while keeping the same person or character.

The Reference Only feature is one of the most amazing technologies in ControlNet.

To use Reference Only, drop an image into the ControlNet section, enable options, and select Reference Only from Preprocessor.

If Reference Only is not visible, an update to ControlNet is required.

The generated image with Reference Only will look very similar to the original dropped image.

Batch count can be increased for more variations while maintaining similarity in faces.

Control Modes can be changed for different outcomes, such as prioritizing the prompt over the reference image.

Focusing on ControlNet with the right photo can yield impressive results.

Adding prompts like 'Running' while keeping the same face demonstrates the versatility of ControlNet.

ControlNet can be used to create cartoons, animations, and frame-by-frame GIF movies.

The XYZ Plot section in Scripts offers useful features for applying prompts to different models.

Illustrative images are compatible with Reference Only, facilitating detailed depictions in cartoons and animations.

Changes such as hair color can be made while keeping the character's face consistent.

Increasing the number of batches and careful selection can improve the accuracy of generated images.

Specific instructions can be given, like 'hold an apple in the girl's right hand', for more accurate results.

Multi ControlNet allows for the application of multiple controlnets to find suitable images for certain prompts.

Openpose settings can be added from ControlNet to adjust poses in generated images.

Adjusting Control Weight and Mode can help retain the original face while changing other elements.

The 'Reference Only' feature significantly lowers the barrier for generating ideal images.

There are many more applications for Reference Only, encouraging users to explore and experiment.

Casual Browsing

学習不要で同じ顔を作れるControlNetの新機能「Reference-only」がすごすぎる【Reference Adain+attnも解説！】

2024-04-14 09:10:06

同じ顔・キャラクターを使ってポーズや表情を指定できるSeaArtの『Reference Only』を使ってみた。

2024-03-27 15:45:02

WebUI ForgeでPhotomakerを使う方法を紹介します【Stable Diffusionで画像と同じ人物を生成する！】

2024-04-16 14:20:02

【凄すぎ！】このAIサイトが凄すぎる！【絶対使え！】

2024-03-30 14:40:01

Stable Diffusionの新機能『IP Adapter』でトレースが可能に。コントロールの超おすすめ機能

2024-04-21 13:35:01

【誰でもできる！】ChatGPTとMidjourneyで超リアルな画像生成をする方法。

2024-04-16 02:00:01