spine-pixi Runtime Documentation


See the Spine Runtimes License before integrating the Spine Runtimes into your applications.

Getting Started

The spine-pixi runtime is implemented on top of spine-ts core, a TypeScript implementation of the renderer-agnostic Spine Runtimes core APIs. The spine-pixi runtime is currently compatible with PixiJS 7, using WebGL to render. Rendering via the canvas APIs is not supported.

The spine-pixi runtime supports all Spine features.


To use spine-pixi in your Pixi project, you must first include its sources.

Vanilla JavaScript

In vanilla JavaScript, use a script tag to include the spine-pixi runtime from unpkg (or host it yourself):

<script src="https://unpkg.com/@esotericsoftware/spine-pixi@4.2.*/dist/iife/spine-pixi.js"></script>

Note: Ensure that the major.minor version of spine-pixi matches the major.minor Spine Editor version you are exporting from. See Synchronizing versions for more information.

The Spine extension will be auto-installed and the spine-pixi runtime can now be used in your Pixi project. Check out index.html for a full example.

The spine-pixi package provides source maps for debugging, as well as minified versions of spine-pixi, which can be used by replacing the .js file suffix with .min.js in the unpkg URLs.

To build spine-pixi.js yourself, follow the instructions in the spine-ts README.md.

NPM or Yarn

When using NPM or Yarn for dependency management, add spine-pixi the usual way:

npm install @esotericsoftware/spine-pixi@~4.2.0

Note: Ensure that the major.minor version of spine-pixi matches the major.minor Spine Editor version you are exporting from. See Synchronizing versions for more information.

Next, just import the Spine class.

import PIXI from "pixi.js"
import { Spine } from '@esotericsoftware/spine-pixi';

The Spine extension will be auto-installed and the spine-pixi runtime can now be used in your project. Check out the spine-pixi esbuild/TypeScript project for a minimal example.

Our module packages contain source maps as well as d.ts typings for improved debugging and development.


The spine-pixi runtime includes several examples demonstrating its feature set.

To run the examples locally:

  1. Install Git and Node.js for your operating system.
  2. Clone the spine-runtimes repository: git clone https://github.com/esotericsoftware/spine-runtimes
  3. Navigate to spine-runtimes/spine-ts, then run npm install & npm run dev.

This builds the spine-pixi runtime, then opens a browser, displaying the example index for all spine-ts based runtimes.

Click on the spine-pixi example you are interested in and check out the code in the spine-runtimes/spine-ts/spine-pixi/example folder.

Updating the spine-pixi Runtime

Before updating your project's spine-pixi runtime, consult our guide on Spine editor and runtime version management.

To update the spine-pixi runtime in vanilla JavaScript, change the version string in the src attribute or the script tag fetching spine-pixi from unpkg.

To update the spine-pixi runtime when managing dependencies with NPM or Yarn, change the version string in your package.json file.

Note: If you change the major.minor version of the spine-pixi package, you have to re-export your Spine skeletons with the same Spine Editor major.minor version. See Synchronizing versions for more information.

Using spine-pixi

The spine-pixi runtime supports all Spine features. It uses the WebGL renderer, while the Canvas renderer is not supported.

Asset Management

Exporting for spine-pixi

Follow the instructions in the Spine User Guide on how to:

  1. Export skeleton & animation data
  2. Export texture atlases containing the images for your skeletons

An export of the skeleton data and texture atlas will yield the following files:

  1. skeleton-name.json or skeleton-name.skel, containing your skeleton and animation data, either in the JSON or binary format.
  2. skeleton-name.atlas, containing information about the texture atlas.
  3. One or more .png files, each representing a page in your texture atlas containing the packed images your skeleton uses.

Note: You should prefer binary skeleton exports over JSON exports, as they are smaller in size and faster to load.

When serving these files, make sure the server emits the correct MIME types:

  • skel files as application/octet-stream
  • json files as application/json
  • atlas files as application/octet-stream
  • png files as image/png

Updating Spine Assets

During development, you may frequently update your Spine skeleton data and texture atlas files. You can simply overwrite these source files (.json, .skel, .atlas, .png) by re-exporting from the Spine Editor and replacing the existing files in your Pixi project.

Ensure that the major.minor version of spine-pixi matches the major.minor Spine Editor version you are exporting from. See Synchronizing versions for more information.

Core classes

The spine-pixi API is built on top of the generic TypeScript spine-core runtime, which provides platform independent core classes and algorithms to load, query, modify, and animate Spine skeletons.

Here, we will briefly discuss the most important core classes that you will encounter in your day-to-day use of spine-pixi. Consult the Spine Runtimes Guide for a detailed overview of the Spine Runtimes architecture, core classes, and API usage.

The TextureAtlas class stores the data loaded from an .atlas file and its corresponding .png image files.

The SkeletonData class stores the data loaded from a .json or .skel skeleton file. The skeleton data contains information about the bone hierarchy, slots, attachments, constraints, skins, and animations. A SkeletonData instance is usually loaded by providing an Atlas from which it sources the images to be used by the skeleton it represents. It serves as a blueprint for creating Skeleton instances. Multiple skeletons can be instantiated from the same atlas and skeleton data, which then share the loaded data, minimizing both load times and memory consumption at runtime.

The Skeleton class stores an instance of a skeleton, created from a SkeletonData instance. A skeleton stores its current pose, that is the position of bones and the current configuration of slots, attachments, and active skin. The current pose can be computed by either manually modifying the bone transforms, or, more commonly, by applying animations via an AnimationState.

The AnimationState class is responsible for keeping track of which animation(s) should be applied to a skeleton, advancing and mixing those animations based on the elapsed time between the last and current rendering frame, and applying the animations to a skeleton instance, thereby setting its current pose. The AnimationState queries an AnimationStateData instance to retrieve mixing times between animations, or fetches the default mix duration if no mix duration is available for a pair of animations.

The spine-pixi runtime builds on top of these core classes.

Spine Pixi runtime

The spine-pixi runtime automatically installs two extensions of type LoadParser into Pixi: skeletonLoader and atlasLoader. They add to PIXI.Assets the functionality to (pre-)load exported .json, .skel, and .atlas files.

The Spine class extends the Pixi Container class and provides a factory function to create Spine container instances from loaded skeleton data and atlas files.

Additionally, it installs an extension of type RendererPlugin that is used to render Spine containers that use tint black on at least one attachment.

Loading Spine Assets

Spine assets, like skeleton data .json/.skel files or .atlas files, are loaded through the usual functions available in the PIXI.Assets class instance, such as Assets.load.

Before an instance of a Spine container can be created, the respective skeleton and atlas files must be loaded. One way of doing it is through the Assets.add and Assets.load functions.

  • Assets.add({ alias: string, src: string }): allows to specify how to resolve the asset alias using the url. This function can be used for all spine assets files (.json, .skel, and .atlas).
  • Assets.load(string[]): loads the asset alias previously added using Assets.add.

Assuming you have exported your skeleton data to a binary skeleton file called skeleton.skel, and your atlas to a file called skeleton.atlas with one corresponding skeleton.png file, you can load your assets like this:

PIXI.Assets.add({ alias: "skeleton-data", src: "path/to/skeleton.skel" });
PIXI.Assets.add({ alias: "skeleton-atlas", src: "path/to/skeleton.atlas" });
await PIXI.Assets.load["skeleton-data", "skeleton-atlas"];

The Assets.load function loads the SkeletonData from the skeleton.skel file and caches it under the key skeleton-data. It also loads the TextureAtlas from the skeleton.atlas file, as well as a texture from the corresponding skeleton.png file. The atlas is cached under the key skeleton-atlas. The individual texture atlas page images are loaded transparently without the need to explicitly load them.

Once preloading has finished, you can access the TextureAtlas via Asset.get(atlasKey). Similarly, you can access the raw .skel file via Asset.get(skeletonKey). Note that in this phase the SkeletonData instance is not available yet.

The raw skeleton data and atlas on their own can not be animated or rendered. Instead, a Spine container is constructed from them. Spine containers that are instantiated with the same asset key share the same skeleton data and atlas.

Creating Spine container instances

Once raw skeleton data and a corresponding atlas have been loaded, a Spine container can be created via the from() static function from the Spine class:

// Pixi app creation
const app = new PIXI.Application({ ... });
// Create a Spine container through the Spine.from factory
const spineboy = Spine.from("spineboyData", "spineboyAtlas", { ... });

// Add the Spine container to the stage

The from() function on the Spine class takes the key of the skeleton data, the key of the atlas, and a ISpineOptions as a parameter.

By default, Spine container bounds are not calculated. You can call getBounds() function to calculate bounds that are sized based on the current skin and animation. If you want the bounds based based on its setup pose, call getBounds() right after the container creation. You may want to configure skins first.

Spine Container

A Spine container is an extension of a Pixi Container that handles storing, updating, and rendering a Skeleton and its associated AnimationState. Spine container instances are created from a skeleton data and an atlas, as described in the last section. The Skeleton and AnimationState are accessible through the skeleton and state fields respectively.

Every frame, the Spine container will:

  • Update the AnimationState
  • Apply the AnimationState to the Skeleton
  • Update the Skeleton world transforms, resulting in a new pose
  • Render the Skeleton in its current pose

Applying Animations

Applying animations to a skeleton displayed by a Spine container is done using AnimationState.

Note: See Applying Animations in the Spine Runtimes Guide for more in-depth information, specifically about animation tracks and animation queueing.

To set a specific animation on track 0, call AnimationState setAnimation:

spineObject.state.setAnimation(0, "walk", true);

The first parameter specifies the track, the second parameter is the name of the animation, and the third parameter defines whether to loop the animation.

You can queue multiple animations using addAnimation:

spineObject.state.setAnimation(0, "walk", true);
spineObject.state.addAnimation(0, "jump", 2, false);
spineObject.state.addAnimation(0, "run", 0, true);

The first parameter to addAnimation is the track. The second parameter is the name of the animation. The third parameter specifies the delay in seconds, after which this animation should replace the previous animation on the track. The final parameter defines whether to loop the animation.

In the example above, the "walk" animation is played back first. 2 seconds later, the "jump" animation is played back once, followed by a transition to the "run" animation, which will be looped.

When transitioning from one animation to another, AnimationState will mix (crossfade) the animations for a specific duration. These mix durations are defined in an AnimationStateData instance, from which the AnimationState retrieves them.

The AnimationStateData instance is also available through the AnimationState.data property. You can set the default mix duration, or the mix duration for a specific pair of animations:

spineObject.state.data.setDefaultMix = 0.2;
spineObject.state.data.setMix("walk", "jump", 0.1);

When setting or adding an animation, a TrackEntry object is returned, which allows further modification of that animation's playback. For example, you can set the mix duration or reverse the animation playback:

const entry = spineObject.state.setAnimation(0, "walk", true);
entry.mixDuration = 0.4;
entry.reverse = true;

See the TrackEntry class documentation for more options.

Note: Be careful about holding on to TrackEntry instances outside the function you are using them in. Track entries are re-used internally and will thus become invalid once the track entry dispose event occurs.

You can use empty animations to smoothly mix the skeleton from the setup pose to an animation, or from an animation to the setup pose:

spineObject.state.setEmptyAnimation(0, 0);
spineObject.state.addAnimation(0, "walk", 0).mixDuration = 0.5;
spineObject.state.addEmptyAnimation(0, 0.5, 6);

Like setAnimation, the first parameter to setEmptyAnimation() specifies the track. The second parameter specifies the mix duration in seconds used to mix out the previous animation and mix in the "empty" animation.

Like addAnimation, the first parameter to addEmptyAnimation() specifies the track. The second parameter specifies the mix duration. The third parameter is the delay in seconds, after which the empty animation should replace the previous animation on the track via mixing.

All animations on a track can be cleared immediately via AnimationState.clearTrack(). To clear all tracks at once, AnimationState.clearTracks() can be used. This will leave the skeleton in the last pose it was in, which is not usually desired. Instead, use empty animations to mix smoothly to the setup pose.

To reset a skeleton to its setup pose, use Skeleton.setToSetupPose():


This will reset both the bones and slots to their setup pose configuration. Use Skeleton.setBonesToSetupPose() or Skeleton.setSlotsToSetupPose() to only reset the bones or slots to their setup pose configuration.

AnimationState Events

An AnimationState emits events during the lifecycle of an animation that is being played back. You can listen for these events and react as needed. The Spine Runtimes API defines the following event types:

  • start: emitted when an animation is started.
  • interrupt: emitted when an animation's track was cleared, or a new animation was set.
  • end: emitted when an animation will never be applied again.
  • dispose: emitted when the animation's track entry is disposed.
  • complete: emitted when an animation completes a loop.
  • event: emitted when a user defined event happened.

To receive events, you can register an AnimationStateListener callback with either the AnimationState to receive events across all animations, or with the TrackEntry of a specific animation queued for playback:

   start: (entry) => log(`Started animation ${entry.animation.name}`),
   interrupt: (entry) => log(`Interrupted animation ${entry.animation.name}`),
   end: (entry) => log(`Ended animation ${entry.animation.name}`),
   dispose: (entry) => log(`Disposed animation ${entry.animation.name}`),
   complete: (entry) => log(`Completed animation ${entry.animation.name}`),
   event: (entry, event) => log(`Custom event for ${entry.animation.name}: ${event.data.name}`)         

trackEntry.listener = {
   event: (entry, event) => log(`Custom event for ${entry.animation.name}: ${event.data.name}`)

See the events-example.html example.


Many applications and games allow users to create custom avatars out of many individual items, such as hair, eyes, pants, or accessories like earrings or bags. With Spine, this can be achieved by using skins.

You can create custom skins from other skins like this:

const skeletonData = spineObject.skeleton.data;
const skin = new spine.Skin("custom");

Create a new, empty skin with the Skin() constructor.

Next, fetch the SkeletonData from the skeleton. It is used to look up skins by name via SkeletonData.findSkin().

Add all the skins you want to combine into the new skin via Skin.addSkin().

Finally, set the new skin on the Skeleton and call Skeleton.setSlotsToSetupPose() to ensure no attachments from previous skins and/or animations are left attached.

See mix-and-match-example.html for full example code.

Setting Bone Transforms

When authoring a skeleton in the Spine Editor, the skeleton is defined in what is called the skeleton's world coordinate system or "skeleton coordinate system". This coordinate system may not align with the coordinate system of Pixi. Mouse and touch coordinates relative to the Spine container need to be converted to the skeleton coordinate system, for example if a user should be able to move a bone by touch.

The Spine container offers the method pixiWorldCoordinatesToBone(point: { x: number, y: number}, bone: Bone) which takes a point relative to the Spine container and converts it to the skeleton's coordinate system, relative to the specified bone.

The reverse, that is converting from the skeleton coordinate system to the Pixi coordinate system, can be achieved via Spine.skeletonToPixiWorldCoordinates(point: { x: number, y: number}).

See control-bones-example.html for full example code.

Adding Pixi Objects to Slots

The Spine class has three convenient methods to attach and detach pixi Containers to slots.

addSlotObject (slotRef: number | string | Slot, pixiObject: Container): void

This adds the pixiObject to the slot referenced by slotRef. You can pass either the name, the index of the slot, or the slot object itself.

It is possible to assign only one Pixi object per slot. Once the Pixi object has been added, its transform will be automatically modified by the Spine object.

To have more control over the added objects, you can add a Pixi Container where you are free to add as many Pixi objects as you want, offset their position, angle, scale, and so on.

If you want to remove a Container, call the appropriate method:

removeSlotObject (slotRef: number | string | Slot, pixiObject?: Container): void

pixiObject is optional. If passed, the Pixi object will be removed only if it corresponds to the one in the slot referenced by slotRef.

Note that the Pixi object is only removed, not destroyed. You should take care of the lifecycle of the Pixi objects added manually.

To retrieve a Pixi object attached to a slot, use the method:

getSlotObject (slotRef: number | string | Slot): Container | undefined

It returns the Container attached to the slot referenced by slotRef, if any.

See slot-objects.html for full example code.

Mesh batch size

Spine objects use Pixi meshes to render attachments. By default in Pixi, a mesh having a number of vertices higher than 100 is marked as not batchable. This will consequently break batching, increasing the number of draw calls and reducing performances. To overcome this inconvenience, you can set the global PIXI.Mesh.BATCHABLE_SIZE to a value that fits your skeletons.

Spine Runtimes API access

spine-pixi exposes the entire spine-ts core API via the Spine properties skeleton, animationStateData, and animationState. See the JS doc documentation of these classes as well as the generic Spine Runtimes Guide.

Differences to pixi-spine

Pixi offers its own Spine extension, also known as pixi-spine, which uses our spine-ts core generic runtime. The main difference is that pixi-spine recreates the Spine skeleton hierarchy as Pixi objects, while spine-pixi is a single Pixi object that uses lower level Pixi Textures and Meshes for more efficient rendering. Note that you can still attach Pixi objects to slots.

Starting with Spine 4.2, our spine-pixi runtime will be maintained along with our other Spine Runtimes. This means you'll get timely updates whenever a new Spine Editor version is released, you will have access to beta runtimes for the next Spine Editor version, and you'll receive all the latest improvements and bug fixes in a timely manner. The Pixi maintainers may not be able to provide such guarantees, as the Spine Extension is understandably not their number one priority.

On the API side, there are a handful of differences to watch out for.

Loading skeleton and atlas data

Loading a skeleton and its atlas with pixi-spine requires atlas and skeleton files having the same name. You just pass the skeleton url to the Assets loader and it will return an object containing the SkeletonData and the TextureAtlas.

const { spineData, spineAtlas } = await Assets.load("spineboy.skel");

With the spine-pixi runtime, the atlas data and skeleton data are loaded separately and can have different names.

await Assets.load(["spineboy-pro.skel", "spineboy-pma.atlas"]);

Creating Spine Container instances

Creating a Spine Container with pixi-spine requires passing the SkeletonData to the Spine constructor.

const spineboy = new Spine(spineData);

The spine-pixi runtime requires you to specify the assets reference of both the skeleton data and atlas.

const spineboy = Spine.from("spineboy-pro.skel", "spineboy-pma.atlas");