Stable Diffusion 1.5 with Fooocus - Yes you can!

Kleebz Tech AI
4 Jun 202408:45

TLDRIn this video, Rodney from Kleebz Tech demonstrates how to use Stable Diffusion 1.5 with Fooocus, despite it being designed for SDXL. He explains the process of downloading a 1.5 model from CivitAI, placing it in the correct folder, and adjusting the refiner switch to blend SDXL and SD 1.5. Rodney also shows how to integrate SD 1.5 compatible LoRAs for specific styles. He encourages viewers to experiment with settings for the best results and invites feedback from the community.

Takeaways

  • 😀 Stable Diffusion 1.5 can be used in Fooocus, although it's primarily designed for SDXL.
  • 🔍 The process of using SD 1.5 in Fooocus is slightly different and may yield different results compared to other user interfaces.
  • 🎨 The base model in Fooocus must be an SDXL model, but the refiner slot can accommodate both SDXL and SD 1.5 models.
  • 📥 To use SD 1.5, download the model from a source like CivitAI, ensuring to select the correct file format, preferably the .safe tensors.
  • 📁 Organize downloaded models by creating a separate folder for SD 1.5 models within the checkpoints directory in the Fooocus models folder.
  • 🔄 Refresh all files in Fooocus to make the newly downloaded SD 1.5 model available for use.
  • 🔄 Use the refiner switch in Fooocus to control at what point during the generation process the SD 1.5 model takes over from the base model.
  • 🎭 Experiment with the refiner switch settings to find the balance between the base model and SD 1.5 that works best for your desired output.
  • 🧩 LoRAs designed for SD 1.5 can be used in Fooocus by downloading them and placing them in the appropriate LoRAs folder within Fooocus.
  • 📸 Generate images using the combination of base model, refiner model, and LoRAs to create unique outputs with Fooocus and SD 1.5.
  • 👍 The video encourages viewers to like, comment with tips, and subscribe to support the channel and explore the creative potential of using SD 1.5 with Fooocus.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is discussing how to use Stable Diffusion 1.5 in Fooocus, despite it being designed for SDXL.

  • What is Rodney's preference for Stable Diffusion models?

    -Rodney personally prefers using SDXL over SD 1.5.

  • Where does Rodney recommend to find SD 1.5 models?

    -Rodney recommends finding SD 1.5 models on CivitAI.

  • What file format does Rodney suggest using for SD 1.5 models in Fooocus?

    -Rodney suggests using the .safe tensors file format for SD 1.5 models in Fooocus.

  • How does one switch from the base model to the refiner model in Fooocus?

    -In Fooocus, you can switch from the base model to the refiner model using the slider provided in the interface.

  • What is the purpose of the refiner switch in Fooocus?

    -The refiner switch in Fooocus determines at what point during the generation process the SD 1.5 model takes over from the base model.

  • How does Rodney demonstrate the use of SD 1.5 with Fooocus?

    -Rodney demonstrates the use of SD 1.5 with Fooocus by generating an image of a cat and adjusting the refiner switch to show the transition.

  • What is the significance of the ribbon logo in the video?

    -The ribbon logo is the trigger word for a specific LoRA that Rodney uses to demonstrate how LoRAs work with SD 1.5 in Fooocus.

  • How does Rodney suggest organizing SD 1.5 models and LoRAs in Fooocus?

    -Rodney suggests creating a separate folder labeled '1.5' to organize SD 1.5 models and LoRAs, making it easier to keep track of them.

  • What is the role of LoRAs in generating images with SD 1.5 in Fooocus?

    -LoRAs in Fooocus with SD 1.5 are used to apply specific styles or effects to the generated images, as demonstrated by the ribbon logo LoRA.

  • How does Rodney encourage the audience to engage with the video?

    -Rodney encourages the audience to like the video, leave comments with tips, and subscribe to the channel.

Outlines

00:00

🖥️ Using Stable Diffusion 1.5 with Fooocus

In this section, Rodney from Kleebz Tech explains how to use Stable Diffusion 1.5 in Fooocus, despite it being designed for SDXL. He clarifies that although the results might differ from using it with other interfaces, it is still functional. Rodney prefers SDXL but acknowledges that using 1.5 allows for the use of certain LoRAs. He guides viewers on how to select the appropriate base model and refiner slot in Fooocus, emphasizing that the refiner slot is where SD 1.5 should be placed. Rodney also provides a step-by-step tutorial on downloading a 1.5 model from CivitAI, choosing the full model in safe tensor format, and organizing it within the Fooocus models folder. He demonstrates how to refresh the model list in Fooocus to include the newly downloaded model and adjust the refiner switch to control the influence of the base model versus the SD 1.5 model during the generation process.

05:01

📸 Generating Images with SD 1.5 and LoRAs in Fooocus

Rodney continues by discussing the use of LoRAs with SD 1.5 in Fooocus. He shows how to download a safe tensor LoRA that is compatible with SD 1.5 and organize it in the LoRAs folder within Fooocus. He explains the importance of refreshing the model list to include the new LoRA and how to select it for use. Rodney then demonstrates the generation process, adjusting the refiner switch to control when the SD 1.5 model takes over during image generation. He emphasizes the need to experiment with the refiner switch to achieve the desired results. Finally, Rodney generates a set of images using a vector art style and a specific LoRA trigger word, showcasing the effectiveness of using SD 1.5 with Fooocus. He concludes by encouraging viewers to like the video, share tips, and support the channel.

Mindmap

Keywords

💡Stable Diffusion 1.5

Stable Diffusion 1.5 is a version of the AI model Stable Diffusion, which is used for generating images from textual descriptions. In the context of the video, it is discussed as a tool that can be integrated with Fooocus, despite it being designed primarily for SDXL. The video explains how to use this model in Fooocus to leverage its capabilities for image generation.

💡Fooocus

Fooocus is a user interface that allows users to interact with AI models like Stable Diffusion for image generation. The video script explains how to configure Fooocus to work with Stable Diffusion 1.5, which is not its default compatibility, showcasing a workaround for using it with certain models and LoRAs.

💡SDXL

SDXL refers to a larger or extended version of the Stable Diffusion model, which is optimized for higher resolution image generation. The video mentions that Fooocus is designed for SDXL, but with specific steps, users can also utilize Stable Diffusion 1.5 within the interface.

💡LoRAs

LoRAs (Low-Rank Adaptations) are modifications or fine-tunings applied to AI models like Stable Diffusion to alter their behavior or specialize them for certain tasks. The script describes how to use LoRAs with Stable Diffusion 1.5 in Fooocus, which can enhance the model's performance for specific image generation tasks.

💡Refiner slot

In the context of the video, the refiner slot is a feature in Fooocus where users can specify a secondary model to be used in conjunction with the base model. It is highlighted as the place where the SD 1.5 model should be loaded to work with Fooocus.

💡CivitAI

CivitAI is mentioned as a preferred source for downloading AI models, including Stable Diffusion 1.5. The video script provides a step-by-step guide on how to find and download a 1.5 model from CivitAI and use it in Fooocus.

💡Checkpoint

A checkpoint in the context of AI models refers to a saved state of the model at a particular stage of training. The video explains that users can choose between different formats of checkpoints, with a recommendation for using the '.safe tensors' format for compatibility with Fooocus.

💡Model Tab

The model tab in Fooocus is where users can select and configure the base model and refiner model for image generation. The video script uses the model tab as part of the demonstration on how to set up Fooocus to work with Stable Diffusion 1.5.

💡Refiner Switch

The refiner switch in Fooocus is a slider that determines at what point during the image generation process the refiner model (in this case, SD 1.5) takes over from the base model. The video provides examples of how adjusting this switch can affect the output of the generated images.

💡Safe Tensors

Safe tensors refer to a specific file format for AI model checkpoints that are considered safer and more compatible with certain interfaces like Fooocus. The video recommends downloading models in this format for use with Fooocus.

Highlights

Stable Diffusion 1.5 can be used in Fooocus despite being designed for SDXL.

Using SD 1.5 in Fooocus will yield slightly different results compared to other UIs.

Rodney prefers SDXL over SD 1.5.

The base model in Fooocus must be an SDXL model.

SD 1.5 can be placed in the refiner slot in Fooocus.

LoRAs for SD 1.5 can be used in the refiner slot as well.

CivitAI is recommended for downloading SD 1.5 models.

Download the full model in .safe tensors format for compatibility with Fooocus.

Create a separate folder for SD 1.5 models in the checkpoints directory for organization.

Refresh all files in Fooocus to detect the newly downloaded SD 1.5 model.

The refiner switch in Fooocus controls the transition from the base model to the refiner model.

Experiment with the refiner switch to find the best results for your needs.

The lower the refiner switch, the more influence SD 1.5 will have on the generation.

Demonstration of generating an image using SD 1.5 in Fooocus.

LoRAs for SD 1.5 must be in safe tensor format to work in Fooocus.

Create a separate folder for SD 1.5 LoRAs in the LoRAs directory for organization.

Refresh all models in Fooocus to detect the newly downloaded SD 1.5 LoRAs.

Using a LoRA with SD 1.5 can result in specific art styles or effects.

The video concludes with a call to action for likes and comments for further tips on using SD 1.5 with Fooocus.