OpenPoser

OpenPoser is a local web-based tool for manually/automatically creating openpose images utilised by ControlNet models.
(Tested with Xinsir's UNION Promax SDXL model - Expects thicker bones).
Simply download and run the HTML file in the .zip file.

Capabilities:

• Human body, hands and animal skeletons

• Multiple skeletons per canvas, with per-skeleton visibility and locking

• Auto-detection per reference - Downloads a small model (20mb) on first run enabling pose auto-detection (Only works with one person per reference)

• Adjustable bone thickness, body, hand, and joint radius

• Joint label overlay toggle

• Movable locked skeletons - Locked skeletons can be moved as a whole

• Multiple reference images - load as many as you need, layer them, and trace poses from each one independently

• Reference image transforms -move, scale (aspect-ratio locked), rotate, flip horizontally/vertically, adjust opacity, lock, and reorder layers

• Pose / Reference mode toggle - switch between posing joints and manipulating reference images without conflicts (press M)

• Custom canvas size with common presets

• Undo / redo - 60-step history

• Zoom controls with Ctrl+scroll support

• Transparent pose-only PNG export (no background - feed directly into ControlNet)

• Export scale multiplier - export at 1×, 2×, 3×, or 4× canvas resolution

• Save/load full editor state as JSON

OpenPoser License

• You may use, copy, and modify this software for personal and non-commercial purposes.

• You may not sell, sublicense, or distribute this software or modified versions of it for profit.

• This software is provided "as is", without warranty of any kind.

• Pose detection uses MediaPipe Holistic by Google (Apache 2.0), loaded on demand from CDN.