Parchment Style with Unsampling

This workflow, created by , explores Unsampling as a method for consistent style transfer in ComfyUI. For a comprehensive guide on using Unsampling, refer to: .

By utilizing this workflow, you can effortlessly convert your video into a parchment style illustration. The process is simple, requiring minimal effort to transform your video seamlessly, resulting in stunning parchment style outputs.

The Unsampling process in ComfyUI transforms the input video into latent noise that retains the original motion and composition. Instead of relying on random noise, this representational noise serves as the starting point for the diffusion process. This allows you to apply the target parchment style while maintaining temporal consistency in the output video.

Key Components of the Parchment Style Unsampling Workflow:

Inputs: Upload your input video here. The frame_load_cap setting limits the number of frames loaded from the input video. The select_every_nth setting skips frames to reduce the total number of frames processed. Increasing the select_every_nth value speeds up processing but may impact the smoothness of the output.

Unsample: The Unsample component uses euler sampling to locate noise that represents the original video. Increasing the cfg scales forces the unsampler to better match the input. The Flip Sigma node reverses the diffusion process, converting the input frames into representative latent noise, critical for the style transfer to parchment.

Prompt: A well-crafted prompt enhances coherence, especially when intensifying the style transfer. For parchment style, a positive prompt could be "ink drawing minimalist illustration on parchment". You can also include more detailed instructions to enhance the texture and intricacy, making the parchment style even more refined and expressive.

ControlNet: The ControlNet component utilizes Depth ControlNet, preprocessed by DepthAnythingV2, to help maintain the structure and geometry of the original video during resampling. The strength setting controls the influence of ControlNet on the output, with higher values adhering more closely to the depth map.

Resample: The Resample component processes the unsampled latent noise by reapplying the diffusion process, this time incorporating the parchment style to enhance the texture and overall visual effect.

Outputs: The decoded frames are saved as images and combined into an upscaled MP4 video at the original frame rate.

The parchment style is just one example of the many creative possibilities available. Feel free to experiment with this Unsampling workflow to achieve your desired aesthetic!