πDocumentation
Welcome to the Anything World API!
You can use the Anything World's API in your application to request 3D models and other data based on different search criteria.
Python users can also use a package directly for easier integration. It's available at https://pypi.org/project/anything-world/ and can be installed with pip install anything-world
.
Basic notions
The API is divided in the library API (used to search and obtain models from our 3D asset library) and the processing API (used to rig, animate or otherwise process your own 3D models using our AI system). However, both use the same API key.
API key
After generating the API key in our Dashboard, the API key can be used to authorize requests to the Anything World's API.
Bear in mind that many endpoints require the API to generate results.
General Format
Example
The API is currently experimental and is subject to change. For use in a critical production environment, please reach out to us at hello@anything.world.
3D Model Library
The following requests can be used to download 3D models from our library. Those are the same models available in our gallery. Here you can find models animated by our AI pipeline as well as static ones, and they are ideal for creating new games or entire 3D worlds without having to design nor animate by hand.
βGet a specific model by name
GET
https://api.anything.world/anything?key=<API_KEY>&name=<NAME>
Retrieve a single model (JSON response) based on its name. If there is an exact match it will be returned, otherwise a similar result (if found), or no model otherwise.
Query Parameters
Get all models matching the query
GET
https://api.anything.world/anything?key=<API_KEY>&search=<QUERY>
Get a JSON array representing all models that matched the query string, which can be anything, such as a name (e.g. "cat"), a tag (e.g. "Christmas"), a habitat (e.g. "jungle" or "desert") or a taxonomical category (e.g. "animal" or "object"). The response is a JSON with an array of objects, where each object represents the data of a model, in the same format as that of the endpoint for obtaining a specific model, shared above.
Query Parameters
AI Model Processing
These requests give you access to our AI pipeline. With them you'll be able to animate your own 3D models. To know which types of models we support and other requirements that should be followed to have a proper animation, please refer to this page.
Automatically rig and animate a given 3D model
POST
https://api.anything.world/animate
Get a JSON response with a "model_id"
representing an unique identifier for the 3D model being animated. Itβs important to note that, because weβre sending files through the request, all parameters and files should be encoded as formdata
.
The files should be sent as the value of the files
key, no matter the amount of files. As value, a list of tuples should be provided with each tuple containing: (file_name, file_content, content_type)
. The file names should have the expected extension and content-type. Currently we support the following file types:
Meshes: OBJ, GLB, GLTF, FBX and DAE
Materials: MTL
Textures: JPG, PNG, BMP, GIF, TIF, TGA and TARGA
As content-type, please make sure your files have the expected one:
Make sure that all your mesh, materials and texture files are referencing each other as expected. For example, a material file named "foo.mtl"
should be properly referenced inside a "model.obj"
. The rule of thumb is: if you are able to open your model in a 3D editor like Blender and it shows all expected materials and textures bindings, it should work with our API as well.
It's also possible to send a ZIP file as your model data. Please make sure to include all the files in a single folder (please, don't use subfolders e.g. for materials, etc) and compress it as a valid .zip file (with the right content-type).
This request is asynchronous and follows a polling strategy, meaning that a response will be returned by the endpoint as fast as possible (in a few seconds). The animation process though takes time (approx. 10 minutes). Please, use the generated "model_id"
as parameter for a request to https://api.anything.world/user-processed-model endpoint to get the current status of the model being animated. When the request returns the final stage, the model should have been animated with success and the generated rigged files will be available to download. Please check the documentation for the https://api.anything.world/user-processed-model endpoint, it's just bellow, in the next section.
To better illustrate this process, we provide an open source Python library that can be used as example on how to communicate with our API: https://github.com/anythingworld/anything-world-python. Given it has minimal requirements, this implementation can be easily translated to other programming languages.
It's important to note that you have a free allowance of model processing credits which renews every month. Additional credits can be purchased in your profile page.
Request Body
Automatically rig a given 3D model (not including animations)
POST
https://api.anything.world/rig
This endpoint is pretty similar to the /animate
endpoint above, however it will only generate the rigging of the provided 3D model instead of also generating all the animations.
Get a JSON response with a "model_id"
representing an unique identifier for the 3D model being rigged. Itβs important to note that, because weβre sending files through the request, all parameters and files should be encoded as formdata
.
The files should be sent as the value of the files
key, no matter the amount of files. As value, a list of tuples should be provided with each tuple containing: (file_name, file_content, content_type)
. The file names should have the expected extension and content-type. Currently we support the following file types:
Meshes: OBJ, GLB, GLTF, FBX and DAE
Materials: MTL
Textures: JPG, PNG, BMP, GIF, TIF, TGA and TARGA
As content-type, please make sure your files have the expected one:
Make sure that all your mesh, materials and texture files are referencing each other as expected. For example, a material file named "foo.mtl"
should be properly referenced inside a "model.obj"
. The rule of thumb is: if you are able to open your model in a 3D editor like Blender and it shows all expected materials and textures bindings, it should work with our API as well.
It's also possible to send a ZIP file as your model data. Please make sure to include all the files in a single folder (please, don't use subfolders e.g. for materials, etc) and compress it as a valid .zip file (with the right content-type).
This request is asynchronous and follows a polling strategy, meaning that a response will be returned by the endpoint as fast as possible (in a few seconds). The rigging process though takes time (approx. 3 minutes). Please, use the generated "model_id"
as parameter for a request to https://api.anything.world/user-processed-model endpoint to get the current status of the model being rigged. When the request returns the final stage, the model should have been rigged with success and the generated rigged files will be available to download. Please check the documentation for the https://api.anything.world/user-processed-model endpoint, it's just bellow, in the next section.
To better illustrate this process, we provide an open source Python library that can be used as example on how to communicate with our API: https://github.com/anythingworld/anything-world-python. Given it has minimal requirements, this implementation can be easily translated to other programming languages.
It's important to note that you have a free allowance of model processing credits which renews every month. Additional credits can be purchased in your profile page.
Request Body
Request Response
OK Response. Created the model and returns the id for it.
Get a model state processed by the AI pipeline
GET
https://api.anything.world/user-processed-model?key=<API_KEY>&id=<MODEL_ID>&stage=<MODEL_STAGE>
Get a JSON response representing the current state of a model being processed by the AI pipeline. If a model has completed the stage specified in the query (or is fully processed if the stage is not given), the response will be a model entry in similar format as the returned JSON of https://api.anything.world/anything endpoint. An error is returned if the model faced a processing error or if the model has not reached yet the asked stage or is past it.
Note that you can speed up your request by setting the stage
parameter in a strategic manner, depending on your use case:
If you don't need other formats than the basic ones provided, you don't need to specify a
stage
parameter (or specifystage="done"
). This way you'll receive a.obj
file for static and vehicle models; and.fbx
and.glb
for animated models. Our AI pipeline will return the models as soon as they are processed, however it will keep performing formats conversion and thumbnails generation on background, and the stage of the model will change forformats_conversion_finished
andthumbnails_generation_finished
when those operations are done. This is the preferred way to get a result as fast as possible;If you care about having the results in
.fbx
and.glb
for static and vehicle models as well as.dae
and.gltf
formats for all models, please request withstage="formats_conversion_finished"
. This will take longer to your request to finish, but you'll be able to have extra formats. You will still get some speed up by not having to wait for thumbnails generation to finish;If you want to receive not only the extra formats but also thumbnail images representing previews of the 3D model, please use
stage="thumbnails_generation_finished"
. This is the longest request and will not give you any speed up when compared with the other previous options.
Within the JSON structure, the model rig sub-object can be found in model.rig
. At its raw level, you can find the rigged mesh file in the formats cited above, e.g. model.rig.FBX
contains the FBX version of the rig. The animations can be found in model.rig.animations
, with a sub-object in the JSON for each animation, and inside that, a field for each format, e.g. model.rig.animations.walk.FBX
contains the walk animation (only available for some categories) in FBX format. If the model is not rigged and animated but a vehicle, you can find the separated parts in the model.parts
sub-object in the JSON response.
It's important to note that you have a free allowance of model processing credits which renews every month. Additional credits can be purchased in your profile page.
Query Parameters
Get all models processed by the AI pipeline for the user
GET
https://api.anything.world/user-processed-models?key=<API_KEY>
Get a JSON response with an array of data from all the models processed with the AI pipeline by the user.
Each element of the array follows the same format of a processed model (rigged and animated model if the category allowed, otherwise part-split vehicle or static mesh) as described in the endpoint user-processed-model
(which you can check for more context, just note that in the endpoint described now you do not have the further control of the stage
parameter).
Within the JSON structure of an element of the array (that is, a processed model), the rig data sub-object can be found in model.rig
. At its raw level, you can find the rigged mesh file in the formats cited above, e.g. model.rig.FBX
contains the FBX version of the rig. The animations can be found in model.rig.animations
, with a sub-object in the JSON for each animation, and inside that, a field for each format, e.g. model.rig.animations.walk.FBX
contains the walk animation (only available for some categories) in FBX format. If the model is not rigged and animated but a vehicle, you can find the separated parts in the model.parts
sub-object in the JSON response.
It's important to note that you have a free allowance of model processing credits which renews every month. Additional credits can be purchased in your profile page.
Query Parameters
Last updated