Y
Prompt Workshop
Article · ControlNet

ControlNet OpenPose Complete Tutorial

OpenPose is ControlNet's most-used preprocessor. It locks a character's pose with a skeleton point graph, and OpenPose Full / DWPose extend to hands and face.

Three OpenPose variants

VariantExtraction scopeBest for
OpenPose (base)Body skeletonBig actions
OpenPose FullBody + fingers + facePrecise hand poses
DWPoseWhole body, more accurateRecommended over Full

Standard workflow

  1. Find a pose reference (real photo or 3D pose tool render)
  2. Preprocessor: DWPose (fallback to OpenPose Full if unavailable)
  3. Verify the extracted skeleton preview looks correct
  4. Weight 0.7–0.9
  5. End step 0.7–0.8 (let the model finish freely)
  6. Write the prompt and generate

Use cases

Lock a specific actionRun, jump, dance

Action reference + weight 0.85.

Lock hand poseGrip, point, hold

DWPose is 10× more accurate than describing the gesture in text.

Two-character compositionConversation, embrace

A 2-person reference yields both skeletons simultaneously.

Pitfalls

Missing skeleton preview

If the preview drops a limb, the generation likely will too. Pick a clearer reference.

Weight 1.0+ rigidifies pose

Stay 0.7–0.9.

Frequently asked questions

Can I use 3D tools for reference?

Yes. Blender, PoseMyArt and similar tools generate pose sketches that work better than photo references.

How much better is DWPose vs OpenPose?

30–50% more accurate at hands and face. Always prefer DWPose if available.

Try this parameter live in the editor

The local editor includes Midjourney / SDXL / Flux templates with a visible, editable parameter panel.

Open the editor →
Yan · AI Prompt Workshop editorial team|Last updated on 2026-06-12。This site does not call any cloud model. Every prompt and parameter in this article was tested and refined locally by the editorial team.