CommunityImagemgithub.com

moho-technique-images

Shows an agent how to direct or produce "moho technique images" so the concrete output is a finished video cut. The procedure moves through Images in Moho, Facts from the manual, Workflow A — parts already separated (works today). Use it when a new assignment calls for this craft.

O que é moho-technique-images?

moho-technique-images is a Claude Code agent skill that shows an agent how to direct or produce "moho technique images" so the concrete output is a finished video cut. The procedure moves through Images in Moho, Facts from the manual, Workflow A — parts already separated (works today). Use it when a new assignment calls for this craft.

Funciona com~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/LuuHung247/moho-studio-mcp/tree/HEAD/skills/moho-technique-images

Installed? Explore more Imagem skills: steipete/songsee, affaan-m/frontend-design, mattpocock/improve-codebase-architecture · View all 6 →

Perguntar na sua IA favorita

Abre um novo chat com esta habilidade de agente já pré-carregada.

Documentação

Images in Moho

Source: Moho 13.5 Tutorials 4.1–4.3. Parts marked (our recommendation) are NOT from the manual. [API] = MCP tool; [GUI] = Moho window.

Facts from the manual

  • An image layer holds ONE image file (or a movie). Use it as a background or inside a bone layer to build a character. Image and vector layers mix freely.
  • Use PNG with an alpha channel (transparent background); non-rectangular parts need transparency. Make it in an image editor — Moho cannot cut images.
  • A new image layer is centered in the workspace: position it afterwards with Transform Layer → [API] layer_setTransform (transX/transY, scaleX/scaleY, rotation). Leave a little overlap between neighboring parts (e.g. at the elbow).
  • Attach a part to a skeleton by binding the layer to a bone (Bind Layer) → [API] layer_setParentBone / rig_attachLayers mode parent.
  • Image warping: Layer Settings → Image tab → check Warp using bones [GUI]. Then the skeleton distorts the bitmap; the nearest bone influences each area most. To keep an area still, add an anchor bone there and never animate it. Best practice: break the picture into logical parts (arms, legs, body as separate images) with separate skeletons/bone layers so a leg bone cannot affect an arm.
  • Textures: render a preview, copy it into an image editor, paint details on their OWN layer, export that layer as PNG and import it as an image layer on top (bound to the same bones).

Workflow A — parts already separated (works today)

  1. Prepare PNGs with alpha: head.png, torso.png, upper_arm_L.png, … (same canvas scale; consistent line style).
  2. layer_createLayer group "Character"; for each part:
{"filePath": "D:\\Phim\\assets\\char\\head.png"}

(returns layerId; name it with layer_setName.) 3. Position/scale each part so it assembles correctly (layer_setTransform), then screenshot and fix overlaps and stacking (layer_placeBehind). 4. Joint positions: measure from the assembled screenshot (units: x −1.78…1.78, y −1…1). Build the rig to MATCH the image, not the other way: rig_buildHumanoid with scale,x,y set to the character's size and hip position, or rig_buildSkeleton with explicit bone positions; then fix joints with bone_setRestPose. 5. Put the parts inside the bone layer (layer_placeInGroup) and bind rigidly: rig_attachLayers mode parent. 6. Animate with anim_* recipes; check joints in a screenshot at an extreme pose and cover gaps with overlap.

{"layerId": 31, "frame": 1, "transX": 0.12, "transY": 0.35, "scaleX": 0.45, "scaleY": 0.45}

Workflow B — one flat character image (AI-generated or photo)

The manual does not cover cutting a picture into parts. Honest options, best first:

  1. Re-generate as parts (recommendation): prompt an image generator for the character in a neutral front pose with arms and legs separated from the body, plain flat background, and ask for a parts sheet (head, torso, upper/lower arms, hands, thighs, shins, feet, eyes, mouth shapes). Remove the background (rembg or chroma-key with ffmpeg) and crop each part to its own PNG with a script in the shell, then Workflow A.
  2. Warp the whole image with bones (manual technique, Moho Pro): import the image, put it in a bone layer, add a few bones along the body/limbs plus anchor bones, enable Warp using bones [GUI]. Gives a floppy "puppet photo" look, fine for simple wobble/talking-head, poor for walking.
  3. Hybrid: keep the body as one warped image, add vector eyes/mouth (draw_*, switch layers) on top for expressions and lip-sync (moho-technique-lipsync).
  4. Convert to vector inside Moho (14.5) [GUI]: Paint Bucket → Trace mode creates NEW vector shapes from regions of a bitmap (enable All layers to use image/video layers as boundaries; Avoid strokes / Pixel expansion control overlap). Trace each body part on its own vector layer, then rig as usual (moho-technique-bones-binding). Alternatively vectorize outside (vtracer/potrace → SVG) and project_import the SVG — Moho 14's improved SVG import keeps layers, gradients and masks.
  5. Curvers (Pro) [GUI]: Draw → Create Curver Layer over a tail, tentacle, hair or limb image; add points at the bends and animate the curve — the bitmap bends with it, without splitting it. See moho-whats-new-14. Never promise the user a full articulated rig from one flat picture; offer these choices and show screenshots.

Backgrounds and props

  • Import big backgrounds at ≥ 2× frame width if the camera moves; set depth with transZ (moho-technique-depth-effects).
  • Layered PSD: import (project_import) keeps layers; check document_getLayers.
  • Sprite/image sequences: project_import with asImageSequence: true (moho-technique-frame-by-frame).
  • Replace art later without re-placing: image_setSource {layerId, filePath} keeps transform and bone binding.
{"layerId": 31, "filePath": "D:\\Phim\\assets\\char\\head_v2.png"}
  • Keep source PNGs ≤ 2048 px per part; huge bitmaps slow preview and render.

Checklist

  • PNGs have alpha and are named by role. [ ] Parts assembled and screenshotted at rest. [ ] Rig joints match image joints (bone_setRestPose). [ ] Parts inside the bone layer and bound. [ ] Extreme-pose test: no tears or missing overlap. [ ] document_save.

Habilidades Relacionadas