Advanced Canvas Inpainting Techniques with Invoke for Concept Art | Pro Diffusion Deep Dive

9 Nov 202321:05

TLDRThe video script outlines a detailed process for concept artists using a digital tool to refine and evolve their initial sketches into polished, high-resolution assets. The artist begins with a line art and uses various features such as control nets, soft edge adapters, and IP adapters to add detail and correct inaccuracies in the generated images. The script emphasizes the importance of adjusting settings like InStep percentage and control weights to achieve the desired level of detail and consistency across different sections of the artwork. The artist also incorporates elements from public domain images to add unique touches to the armor design, demonstrating a creative approach to enhancing the final product.


  • ๐ŸŽจ The video focuses on using artwork to evolve a concept into a close-to-final asset, particularly for concept artists working with digital tools like Invoke and Advanced Canvas.
  • ๐Ÿ–Œ๏ธ The process begins with a detailed line art that is passed through various stages of digital enhancement to achieve the final vision.
  • ๐Ÿ“ฑ Artists often start with a rough sketch on a tablet or in Photoshop before moving to more advanced stages of refinement.
  • ๐Ÿ” The video emphasizes the importance of working with higher resolution (sdxl) for better detail capture and output quality.
  • ๐Ÿ”ง The use of control adapters and soft edge control adapters are highlighted for fine-tuning the generation process and maintaining the desired level of detail.
  • ๐Ÿš€ The video demonstrates the iterative process of generation, tweaking, and re-generation to achieve the desired results.
  • ๐Ÿ‘ค It addresses common challenges faced by concept artists, such as the loss of detail in facial features during the first generation.
  • ๐ŸŽญ The speaker shares techniques for dealing with initial generation misses, like manually adding details or adjusting settings for better results.
  • ๐Ÿงฉ The concept of using 'jigsaw puzzle' technique for complex in-painting tasks is introduced for more coherent and unified results.
  • ๐Ÿ”— The video introduces the use of IP (Image Processing) adapters for additional control over the generation process, especially for maintaining consistency in character features.
  • ๐ŸŒธ The video concludes with an example of incorporating public domain images for creative enhancement, showcasing the blending of traditional and modern elements in digital art.

Q & A

  • What is the primary focus of the video?

    -The primary focus of the video is to demonstrate the process of using artwork to evolve a concept into a close-to-final asset, specifically within the context of concept artists working with Invoke and advanced canvas painting techniques.

  • What does the presenter initially start with in the video?

    -The presenter initially starts with a more detailed image that they had generated prior to the video, which serves as a basis for walking through the process of transforming the work into something new.

  • What is the significance of using an SDXL model in the process?

    -The SDXL model is significant because it deals with higher resolution output, allowing for more detailed and sharper images. It is important to match the pre-processed image's resolution with the new target resolution for the best results.

  • How does the presenter address the issue of lost detail in the face during the first generation?

    -The presenter acknowledges the issue of lost detail in the face and suggests going back to adjust settings or draw a thicker line for better pre-processor pickup. However, they choose a different approach by focusing on the core concepts and continuing the generation process with those in mind.

  • What is the purpose of using a Control Net in the process?

    -The purpose of using a Control Net is to maintain consistency and coherence in the generated image, particularly when regenerating specific sections. It helps to ensure that different parts of the image fit together seamlessly.

  • How does the presenter handle the challenge of painting a lower body region that is difficult to prompt?

    -The presenter uses a combination of techniques, including IP adapter and Control Net, to refine the interpretation and generation of the lower body region. They also focus on specific concepts like 'futuristic machine Samurai' and 'cropped lower body' to guide the generation process.

  • What additional elements does the presenter import from the public domain?

    -The presenter imports a Japanese flower pattern and a Samurai plate with a dragon from the public domain to add unique elements to the artwork.

  • How does the presenter ensure that the imported elements from the public domain integrate well with the existing image?

    -The presenter uses a combination of Control Net, IP adapter, and adjusting prompt weights to ensure that the imported elements like the Japanese flower pattern and the dragon integrate well with the existing image, creating a cohesive final product.

  • What is the presenter's approach to refining and adding details without losing control over the generation process?

    -The presenter's approach involves using a combination of tools and techniques, including Control Net, IP adapter, adjusting prompt weights, and focusing on specific concepts. They also treat the inpainting process like a jigsaw puzzle, mentally segmenting different areas of the image to ensure a unified and consistent result.

  • What is the final outcome of the video?

    -The final outcome of the video is a detailed and refined image of a Samurai with a decorated armor, achieved by using a variety of advanced canvas painting techniques and inputs to guide the generation process towards the presenter's creative vision.

  • What advice does the presenter give for those looking to improve their workflow and achieve interesting concepts?

    -The presenter advises viewers to explore the various tips and tricks demonstrated in the video, such as using Control Net, IP adapter, and different prompting techniques, to control the generation process, guide generations towards their creative vision, and rapidly accelerate their workflows towards near-final assets.



๐ŸŽจ Artwork Evolution and Concept Refinement

The paragraph discusses the process of using artwork to evolve a concept into a near-final asset. It highlights the collaboration between concept artists and the use of software like Invoke to refine rough sketches into detailed visions. The speaker explains their approach to adding detail, focusing on the image, and tweaking ideas to achieve the desired final asset. They also touch on advanced concepts for users familiar with the software and emphasize the importance of resolution and image processing in achieving sharp, detailed outputs.


๐Ÿ–Œ๏ธ Enhancing Details and Addressing Generation Misses

This section delves into the challenges faced by concept artists during the first generation of their work, particularly the loss of detail in areas like the face. The speaker shares techniques for improving these outcomes, such as adjusting settings, refining the line work, and using control adapters to give more freedom to the generation process. They also discuss the importance of selecting the right areas of the image for in-painting and the use of control nets to maintain consistency and coherence in the final image.


๐Ÿ› ๏ธ Advanced Canvas Techniques and Controls

The speaker introduces advanced canvas painting techniques, including the use of IP adapters and control nets for more precise control over the generation process. They discuss the use of different adapter models for concept, position, and face consistency, as well as the importance of articulating the elements within the bounding box for effective generation. The paragraph also explores the creative use of public domain images to add unique elements to the artwork, emphasizing the balance between control and freedom in the generation process.


๐Ÿ” Focusing on Details and Final Touches

In this part, the speaker focuses on perfecting specific areas of the image, such as the dragon detail on the armor, and the overall decoration of the Samurai character. They discuss the iterative process of zooming in on details, using control adapters and IP adapters to refine the image, and the importance of maintaining the character's face consistency. The speaker also shares their determination to make the dragon work and the meticulous approach to enhancing the armor's design, demonstrating a commitment to achieving a high-quality final asset.


๐Ÿš€ Accelerating Workflows and Creative Vision

The conclusion of the video script emphasizes the multitude of tips and tricks available for controlling the generation process and guiding it towards a creative vision. The speaker encourages viewers to share their feedback and expresses eagerness to provide more content in the future. This paragraph serves as a recap of the strategies discussed throughout the video and an invitation for the audience to engage with the content and continue their exploration of the creative process.




Artwork refers to the visual creations made by the concept artists, which are the starting points for evolving a concept into a final asset. In the video, the speaker begins with a detailed artwork and walks through the process of transforming it using various tools and techniques.


Invoke appears to be the name of a software or tool used in the process of taking the rough sketch and refining it. It is where the initial transformation of the artwork happens, moving it closer to the final vision.

๐Ÿ’กControl Nut

Control Nut is a tool or feature within the software that allows the user to manipulate and control specific aspects of the image, such as sharpness and detail. It is used to fine-tune the generated images during the creation process.

๐Ÿ’กHD Processor

HD Processor is a term used to describe a high-definition processing tool that enhances the quality of the pre-processed image, making it sharper and more detailed.


Canvas refers to the workspace or platform where the artist takes the processed output and further refines it, adding more details and addressing core misses from the initial generation.

๐Ÿ’กSoft Edge Control Adapter

Soft Edge Control Adapter is a feature that allows the artist to control the softness or sharpness of the edges in the image. It is used to adjust the level of detail and the overall look of the artwork.

๐Ÿ’กDenoising Strength

Denoising Strength is a parameter that controls the level of noise reduction applied to the image during the generation process. Adjusting this parameter can affect the clarity and detail of the final output.

๐Ÿ’กControl Net

Control Net is a tool that helps in maintaining consistency and coherence in the image by using a reference image to guide the generation process. It is particularly useful for complex or detailed sections of the artwork.

๐Ÿ’กIP Adapter

IP Adapter is a tool that provides additional control over the generation process, allowing the artist to focus on specific concepts and positions within the image. It helps in achieving a more precise and desired outcome.

๐Ÿ’กAdvanced Canvas Painting

Advanced Canvas Painting refers to the techniques and methods used to refine and add intricate details to the artwork using various tools and inputs. It involves a higher level of skill and creativity to achieve a polished and final asset.


Workflow in this context refers to the sequence of steps and processes followed by the artist to transform a rough sketch into a detailed and polished final asset. It encompasses the use of different tools, techniques, and adjustments made throughout the creation process.


The video focuses on breaking down the process of using artwork to evolve a concept into a close-to-final asset, particularly for concept artists.

The process involves taking a rough sketch and refining it using software like Invoke to get closer to the final vision.

The presenter uses a more detailed starting point for the demonstration, which is useful for understanding the transformation process.

The importance of working with higher resolution (sdxl) is emphasized for better detail and sharpness in the final output.

The use of control nets and soft edge control adapters is discussed to add more detail and focus to the image.

The presenter addresses the common issue of losing detail in the first generation, especially in facial features.

A different approach is suggested for refining the generation process, focusing on maintaining core concepts from the initial sketch.

The process of selecting the most interesting coloring and details from generated options is explained.

The presenter demonstrates how to adjust settings and regenerate sections of the image for better detail and resolution.

The use of IP (image processing) adapters is introduced for additional control over the generation process.

The presenter shares a trick for using the unified canvas to regenerate sections like a jigsaw puzzle, ensuring a coherent result.

The video showcases the technique of reimporting the image from the canvas for new snapshots and references.

The presenter discusses the challenges of prompting around certain regions and shares techniques for refining those sections.

The video includes an example of incorporating public domain images into the canvas for unique design elements.

The presenter emphasizes the importance of checking control adapters for accurate outputs during the creative process.

The video concludes with a summary of tips and tricks for controlling the generation process and accelerating workflows towards final assets.