Floyo
Floyo
Workflows
API
Pricing
Floyo
Floyo
Workflows
API
Pricing

h3-latent-upscaler

Created by rockerBOO

https://github.com/rockerBOO/h3-latent-upscaler

11

Last updated 2026-08-05

Last updated

2026-08-05

Run hundreds of ComfyUI nodes and workflows in your browser.

ComfyUI and My Files on Floyo

ComfyUI's h3-latent-upscaler is a set of custom nodes designed to enhance the resolution of MiniMax H3 video latents during processing. This tool facilitates a two-step sampling process that transitions from low-resolution to high-resolution outputs efficiently.

  • Provides specialized nodes for upscaling both video and conditioning inputs using various methods, ensuring optimal quality.
  • Integrates advanced noise handling and sigma shifting to maintain audio and video synchronization, addressing common pitfalls in conventional noise applications.
  • Includes example workflows that demonstrate practical implementations, simplifying the user experience for those working with MiniMax H3 tasks.

Context

The h3-latent-upscaler is a collection of custom nodes within ComfyUI specifically aimed at spatially upscaling video latents generated by the MiniMax H3 model. Its primary purpose is to enhance the quality of video outputs by applying a systematic upscaling method between two sampling passes, allowing for a more refined and detailed final product.

Key Features & Benefits

The tool includes several key nodes that each serve a distinct purpose in the upscaling process. The MiniMax H3 Latent Upscale node allows users to upscale latent video tensors using various interpolation methods, while the MiniMax H3 Conditioning Upscale node ensures that accompanying conditioning inputs are rescaled appropriately to maintain their relative proportions. Additionally, the MiniMax H3 Add Noise (resample) node is designed to replace the standard noise addition process, correcting issues that arise when traditional methods are used with MiniMax H3's unique data structures.

Advanced Functionalities

One of the advanced features of this tool is the MiniMax H3 Shift Sigmas node, which remaps the sigma values used for noise application between video and audio schedules. This functionality is critical for ensuring that the audio quality remains intact and synchronized with the video, addressing discrepancies that can lead to audio distortion. This targeted approach to noise handling and sigma adjustment is essential for achieving high-quality results in complex multimedia projects.

Practical Benefits

By incorporating the h3-latent-upscaler into their workflows, users can significantly enhance the quality and coherence of their video outputs. The tool streamlines the process of transitioning from low-resolution to high-resolution video while maintaining proper audio synchronization. This results in a more efficient workflow, allowing users to achieve better control over their outputs without the need for extensive manual adjustments or troubleshooting.

Credits/Acknowledgments

The h3-latent-upscaler is developed by contributors to the ComfyUI community, with specific implementation details and functionality tailored for the MiniMax H3 model. The project is open source, allowing for collaboration and further enhancements from the community.

Inner Nodes

MiniMaxH3AddNoise
MiniMaxH3ConditioningUpscale
MiniMaxH3LatentUpscale
MiniMaxH3ShiftSigmas

Discover most popular workflows

Hand-picked based on what hundreds of other artists looked at.

Created by rockerBOO

Inner Nodes

MiniMaxH3AddNoise

MiniMaxH3ConditioningUpscale

MiniMaxH3LatentUpscale

MiniMaxH3ShiftSigmas