Repolode
darksidewalker

darksidewalker/ComfyUI-DaSiWa-Nodes

A collection of ComfyUI custom nodes for video generation workflows, resolution control, upscaling, LoRA loading, and prompt building.

Stars
220
Forks
44
Open issues
1
Last commit
2 days ago

What is it?

DaSiWa Nodes is a Python pack of custom nodes for ComfyUI, the node-based Stable Diffusion and video generation interface. Its largest component is the MiniMaxH3 Director, a single node that combines timeline-style reference media (images, videos, audio), prompt editing, RefMod persona files, and continuity extensions for MiniMax H3 video and audio generation.

The collection also includes utility nodes for common video and image tasks: a resolution scale calculator with divisor-aligned presets, a Torch Resize node with Lanczos resampling and chunked batch processing, an Advanced LoRA Loader with a 10-slot stacker, a Seamless Loop node built on RIFE, FILM, or NVIDIA RTX frame generation, and an Enhanced Video Combine node for encoding AV1, VP9, or H.264 output with optional audio. Other nodes cover tiled diffusion for large canvases, inpaint crop and composite, watermarking, metadata-embedding image saving for Civitai, seed control, wildcard prompt building, and a local or remote LLM/VLM prompt writer.

Several nodes depend on NVIDIA RTX SDK features and on MiniMax H3 model components, so they are most useful for users with compatible NVIDIA GPUs and those workflows.

Key features

  • 01MiniMaxH3 Director combines reference media, prompt editing and RefMod control in one node.
  • 02Seamless Loop creates loop transitions with RIFE, FILM or RTX frame generation.
  • 03Enhanced Video Combine encodes AV1, VP9 or H.264 video with optional audio muxing and in-node preview.
  • 04Torch Resize and resolution calculator handle divisor-aligned resizing and megapixel presets.
  • 05Advanced LoRA Loader stacks up to 10 LoRAs with per-slot strength control.
  • 06Metadata Image Saver writes A1111-style parameters and LoRA hashes for Civitai compatibility.

Use cases

  • Generating longer or multi-reference video clips with MiniMax H3 in ComfyUI.
  • Upscaling and sharpening image or video batches while keeping dimensions aligned for a target model.
  • Stacking several LoRAs with individual strength control in a single loader.
  • Publishing generated images to Civitai with prompt and LoRA metadata embedded.

Things to know

  • Licensed under GPL-3.0, a copyleft license that requires derivative works to be distributed under the same terms.
  • Some nodes, such as RTX Upscaler and RTX Frame Generation, require a supported NVIDIA GPU and extra SDK packages.

In collections

Topics

FAQ

What is ComfyUI-DaSiWa-Nodes?
A collection of ComfyUI custom nodes for video generation workflows, resolution control, upscaling, LoRA loading, and prompt building. DaSiWa Nodes is a Python pack of custom nodes for ComfyUI, the node-based Stable Diffusion and video generation interface. Its largest component is the MiniMaxH3 Director, a single node that combines timeline-style reference media (images, videos, audio), prompt editing, RefMod persona files, and continuity extensions for MiniMax H3 video and audio generation.
Is ComfyUI-DaSiWa-Nodes free to use? What license does it have?
ComfyUI-DaSiWa-Nodes is open source under the GPL-3.0 license. Check the license terms before using it in commercial products.
Is ComfyUI-DaSiWa-Nodes still maintained?
Actively maintained. The last commit to ComfyUI-DaSiWa-Nodes was on Nov 5, 2026.
What can I use ComfyUI-DaSiWa-Nodes for?
Generating longer or multi-reference video clips with MiniMax H3 in ComfyUI. Upscaling and sharpening image or video batches while keeping dimensions aligned for a target model. Stacking several LoRAs with individual strength control in a single loader. Publishing generated images to Civitai with prompt and LoRA metadata embedded.
Where can I get ComfyUI-DaSiWa-Nodes?
The source code and installation instructions are on GitHub: https://github.com/darksidewalker/ComfyUI-DaSiWa-Nodes

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects