DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to refine small faces in a MiniMax H3 video with Comfy UI H3 Face Refine

Track each face per frame, crop to it, re-generate it with H3, and stitch it back into the clip.

  • Top pick
  • Content
  • Intermediate
  • 3 hours
  • From the README
  • Any agent

Authorization notice. Comfy UI H3 Face Refine is dual-use. Only use it on software, sites and systems you own or have explicit permission to test.

Who it is forA video creator whose MiniMax H3 clip has smeared faces because the head fills only a small part of the frame.

The problemH3 renders faces poorly when the head is a small fraction of the frame, and that stays true at 720p and above.

The outcomeA composited video where the small face is re-generated from a normalized crop and the rest of the frame keeps its original pixels.

What you need first

  • ComfyUI with MiniMax H3 support (H3 nodes are core in comfy_extras/nodes_minimax_h3.py)
  • A face detector such as face_yolov8m.pt in models/ultralytics/bbox/, which you must supply yourself
  • MiniMax H3 diffusion model, Qwen3-VL text encoder, and the H3 video and audio VAEs in their ComfyUI model folders
  • Git, to clone the pack into ComfyUI/custom_nodes/
  • ComfyUI-VideoHelperSuite if you want to save the output video with VHS_VideoCombine

Steps

  1. Run git clone https://github.com/Carasibana/ComfyUI-H3-FaceRefine.git inside ComfyUI/custom_nodes/.
  2. Restart ComfyUI and confirm the nodes appear under MiniMax H3/Face Refine.
  3. Place face_yolov8m.pt in models/ultralytics/bbox/, and the H3 diffusion model, Qwen3-VL text encoder, and VAEs in their folders.
  4. Load your clip, choose the subject with the select setting, and connect it to H3 Face Track + Crop.
  5. Send the crop through the H3 node, composite the result back, and save the video with a node such as VHS_VideoCombine.

Prompt for your agent

Using the repo Carasibana/ComfyUI-H3-FaceRefine, build a ComfyUI graph that refines the small face of <subject description> in <video path> and writes a composited video to <output path>. Verify that the tracker report shows the subject found on most frames and lists any lost frame ranges.

Caveat. The H3 models and the face detector are not bundled with the pack, so downloading and placing them is the slowest part of setup.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page