ComfyUI Node: Pipeline SuperScaler

Pipeline SuperScaler:

The SuperScaler_Pipeline is a comprehensive node designed to enhance and upscale images through a multi-step process. It combines latent refinement, generative tiled upscaling, and post-processing techniques such as sharpening and adding grain to produce high-quality, detailed images. This node is particularly beneficial for AI artists looking to improve the resolution and visual appeal of their creations without losing the original artistic intent. By integrating several advanced image processing techniques, the SuperScaler_Pipeline offers a streamlined solution for achieving superior image quality, making it an essential tool for those seeking to enhance their digital artwork.

Pipeline SuperScaler Input Parameters:

enable_latent_pass

This parameter is a boolean that determines whether the latent refinement pass is activated. When enabled, it allows the node to perform an initial refinement of the image's latent space, which can enhance the overall quality of the upscaled image. The default value is False.

model_pass_1

This parameter specifies the model used during the latent refinement pass. It is crucial for defining the characteristics and style of the refinement process, impacting the final image's quality and appearance.

vae_pass_1

This parameter refers to the Variational Autoencoder (VAE) used in the latent refinement pass. The VAE is responsible for encoding and decoding the image, playing a vital role in the refinement process.

positive_pass_1

This parameter provides positive conditioning for the latent refinement pass, guiding the model towards desired features or styles in the image.

negative_pass_1

This parameter offers negative conditioning for the latent refinement pass, helping to suppress unwanted features or styles in the image.

latent_upscale_by

This float parameter controls the scale factor for the latent refinement pass, determining how much the image is upscaled during this phase. It ranges from 1.0 to 4.0, with a default value of 1.1.

latent_denoise

This float parameter sets the denoising level during the latent refinement pass, affecting the smoothness and clarity of the refined image. It ranges from 0.0 to 1.0, with a default value of 0.2.

latent_sampler_name

This parameter specifies the sampler used in the latent refinement pass, influencing the sampling strategy and the resulting image's quality.

latent_scheduler

This parameter defines the scheduler used in the latent refinement pass, affecting the timing and sequence of operations during the refinement process.

latent_steps

This integer parameter determines the number of steps taken during the latent refinement pass, impacting the detail and quality of the refined image. It ranges from 1 to 1000, with a default value of 6.

latent_cfg

This float parameter sets the configuration for the latent refinement pass, influencing the balance between fidelity and creativity in the refined image. It ranges from 0.0 to 100.0, with a default value of 1.0.

enable_tiled_pass_2

This boolean parameter activates the second pass of generative tiled upscaling, allowing the node to further enhance the image's resolution and detail. The default value is True.

model_pass_2

This parameter specifies the model used during the second pass of generative tiled upscaling, crucial for defining the characteristics and style of the upscaling process.

vae_pass_2

This parameter refers to the VAE used in the second pass of generative tiled upscaling, responsible for encoding and decoding the image during this phase.

positive_pass_2

This parameter provides positive conditioning for the second pass of generative tiled upscaling, guiding the model towards desired features or styles in the image.

negative_pass_2

This parameter offers negative conditioning for the second pass of generative tiled upscaling, helping to suppress unwanted features or styles in the image.

tiled_upscale_by_2

This float parameter controls the scale factor for the second pass of generative tiled upscaling, determining how much the image is upscaled during this phase. It ranges from 1.0 to 8.0, with a default value of 2.0.

Pipeline SuperScaler Output Parameters:

image_out

The output parameter image_out represents the final upscaled and processed image. This image is the result of the entire SuperScaler_Pipeline process, including latent refinement, generative tiled upscaling, and post-processing. It is delivered as a high-quality, detailed image that maintains the artistic intent of the original while enhancing its resolution and visual appeal.

Pipeline SuperScaler Usage Tips:

• To achieve the best results, start by enabling the latent refinement pass if your initial image requires significant enhancement in detail and quality.
• Experiment with different models and VAEs for each pass to find the combination that best suits your artistic style and desired outcome.
• Adjust the latent_upscale_by and tiled_upscale_by_2 parameters to control the level of upscaling, keeping in mind that higher values may require more processing time.
• Use the positive and negative conditioning parameters to fine-tune the image's features, enhancing desired elements while suppressing unwanted ones.

Pipeline SuperScaler Common Errors and Solutions:

"[SuperScaler] PASS 1: Sauté (désactivé ou inputs manquants)"

• Explanation: This error occurs when the latent refinement pass is either disabled or missing necessary inputs.
• Solution: Ensure that the enable_latent_pass parameter is set to True and that all required inputs, such as model_pass_1 and vae_pass_1, are provided.

"[SuperScaler] PASS 4: Post-processing error"

• Explanation: This error might occur during the post-processing phase, possibly due to incorrect parameter settings or missing inputs.
• Solution: Double-check the post-processing parameters and ensure all necessary inputs are correctly configured. Adjust settings like sharpening and grain to see if the issue resolves.