ONNX Model Zoo

realgen4/OFMHUB-assets

By realgen4 · ↓ 0 · ♥ 0 · Updated 2026-03-21

What's in this repo

This page lists the files in realgen4/OFMHUB-assets, including vitpose_h_wholebody_model.onnx, yolov10m.onnx.

  • 2 ONNX models (.onnx files)
  • 29 files, including 10 .safetensors, 8 .py, 2 .sh, 2 .onnx, 1 .json, 1 .md
  • Last updated 2026-03-21

ONNX files

File Size Download
models/onnx/vitpose_h_wholebody_model.onnx — Download
models/onnx/yolov10m.onnx — Download

How to use

These ONNX files can be loaded with ONNX Runtime. To inspect a model graph, upload the file to the Netron viewer. Download links point to the original files on Hugging Face.

README

ComfyUI Teskor's Utils

Utility nodes for ComfyUI focused on video workflows, OpenPose stability, and chunk-based video generation.

Main use cases:

  • WanVideo
  • AnimateDiff
  • ControlNet OpenPose
  • Long videos
  • Chunked generation

Goal:

Improve stability and visual consistency in generated videos.

Main Nodes

OpenPose Smoother

Purpose

Stabilizes OpenPose pose data across frames.

Raw OpenPose detection is noisy and causes:

  • Pose jitter
  • Hand flicker
  • Missing joints
  • Detection noise
  • Random extra people

What It Does

  • Smooths keypoints over time
  • Fills small gaps
  • Filters unstable detections
  • Optional extra-person removal
  • Outputs smoothed POSEDATA

Typical Pipeline

Load Video
→ OpenPose Detection
→ OpenPose Smoother
→ ControlNet OpenPose
→ Video Generation

Color Match Sequential Bias

Purpose

Removes color drift between video chunks.

Designed for chunk-based generation:

  • WanVideo
  • AnimateDiff
  • Long videos

Problem

Chunked generation causes small differences in:

  • Brightness
  • Color balance
  • Contrast

This makes chunk boundaries visible.

What It Does

For each chunk:

  1. Measures previous chunk colors
  2. Measures current chunk colors
  3. Calculates difference
  4. Applies correction

Result:

  • Consistent colors
  • Invisible chunk borders
  • Continuous video

Important Setting

chunk_size must match generation chunk size.

Example:

If chunks are 81 frames:

chunk_size = 81

Typical Pipeline

WanVideo Animate Embeds
→ Combine Frames
→ Color Match Sequential Bias
→ Save Video

Other Nodes

Additional utilities:

  • Batch video loading
  • Pose data save/load
  • Preview image without metadata
  • Preview video without metadata
  • Batch file renaming

Recommended For

  • WanVideo workflows
  • AnimateDiff workflows
  • OpenPose animation
  • ControlNet OpenPose
  • Long video generation
  • Chunked workflows

Model info

onnx region:us
↓ 0 ♥ 0

Created: 2026-03-21

Updated: 2026-03-21

View on Hugging Face

Files (29)

.gitattributes
.gitignore
OFM_hubv2.json
README.md
__init__.py
animator.sh
lipsync.sh
models/clip_vision/klip_vision.safetensors
models/controlnet/Wan21_Uni3C_controlnet_fp16.safetensors
models/diffusion_models/WanModel.safetensors
models/loras/WanFun.reworked.safetensors
models/loras/WanPusa.safetensors
models/loras/light.safetensors
models/loras/wan.reworked.safetensors
models/onnx/vitpose_h_wholebody_data.bin
models/onnx/vitpose_h_wholebody_model.onnx ONNX
models/onnx/yolov10m.onnx ONNX
models/sam2/sam2.1_hiera_base_plus.safetensors
models/text_encoders/text_enc.safetensors
models/vae/vae.safetensors
nodes/color_match.py
nodes/load_video_batch.py
nodes/openpose_smoother.py
nodes/preview_image_metadata.py
nodes/rename_files.py
nodes/save_load_pose.py
nodes/video_combine_metadata.py
requirements.txt
web/js/ts_video_preview.js

Related models

More from realgen4

No other models by this author.

More models

No task type declared.

Popular models