Skip to main content


This library implements C and C++ APIs that:

This documentation assumes you are using this library with C++.

WARNING: This is an early prerelease version of this library. There may be bugs and unimplemented features, and the API is subject to change.


Add the prebuilt library and header files (include/c2pa.h for C or include/c2pa.hpp for C++) to your project.

For instructions on how to build the library and run the tests and examples, see Development below.


To use this library, include the header file in your code as follows:

#include "c2pa.hpp"

Read and validate C2PA data in a file

Use the read_file function to read C2PA data from the specified file. This function examines the specified asset file for C2PA data and its return value is a JSON report if it finds C2PA data. If there are validation errors, the report includes a validation_status field. Exceptions are thrown on errors.

auto json_store = C2pa::read_file("<ASSET_FILE>", "<DATA_DIR>")


  • <ASSET_FILE>- The asset file to read; The file must be one of the supported file formats.
  • <DATA_DIR> - Optional path to data output directory; If provided, the function extracts any binary resources, such as thumbnails, icons, and C2PA data into that directory. These files are referenced by the identifier fields in the manifest store report.

For example:

auto json_store = C2pa::read_file("work/media_file.jpg", "output/data_dir")

A media file may contain many manifests in a manifest store. The active_manifest property in the manifest store identifies the most recently-added manifest. For a comprehensive reference to the JSON manifest structure, see the CAI manifest store reference.

Create a SignerInfo instance

A SignerInfo object contains information about a signature. To create an instance of SignerInfo, first set up the signer information from the public and private key files. For example, using the simple read_text_file function defined in the training.cpp example:

string certs = read_text_file("path/to/certs.pem").data();
string private_key = read_text_file("path/to/private.key").data();

Then create a new SignerInfo instance using the keys, specifying the signing algorithm in the .alg field and optionally a time stamp authority URL in the .ta_url field:

C2paSignerInfo sign_info = {.alg = "es256", 
.sign_cert = certs.c_str(),
.private_key = private_key.c_str(),
.ta_url = ""};

For the list of supported signing algorithms, see Creating and using an X.509 certificate.

Creating a manifest JSON definition

The manifest JSON string defines the C2PA manifest to add to the file.

A sample JSON manifest is provided in tests/fixtures/training.json.

const std::string manifest_json = R"{
"claim_generator": "c2pa_c_test/0.1",
"claim_generator_info": [
"name": "c2pa-c test",
"version": "0.1"
"assertions": [
"label": "",
"data": {
"entries": {
"c2pa.ai_generative_training": { "use": "notAllowed" },
"c2pa.ai_inference": { "use": "notAllowed" },
"c2pa.ai_training": { "use": "notAllowed" },
"c2pa.data_mining": { "use": "notAllowed" }

Add a signed manifest to a media file

Use the sign_file function to add a signed manifest to a media file.


The parameters are:

  • <ASSET_FILE>- Path to the source (original) asset file to read; The file must be one of the supported file formats.
  • <SIGNED_ASSET_FILE> - Path to the destination asset file that will contain a copy of the source asset with the manifest data added.
  • manifest_json - An string containing a JSON-formatted manifest data to add to the asset; see Creating a manifest JSON definition file.
  • sign_info - A valid SignerInfo object instance; see Generating SignerInfo.
  • <DATA_DIR> - Optional path to data directory from which to load resource files referenced in the manifest JSON; for example, thumbnails, icons, and manifest data for ingredients.

Exceptions will be thrown on errors.

For example:


See training.cpp for an example.


This project has been tested on macOS and should also work on common Linux distributions.


If you haven't already done so, install Rust.

Install cbindgen:

cargo install --force cbindgen


Building the library requires GNU make, which is installed on most macOS systems.

Enter this command to build the C library:

make release

The Makefile also has numerous other targets:

  • test-cpp to build and run the C++ tests.
  • test-c to build and run the C tests.
  • check-format, clippy, and test-rust for Rust linting and testing.
  • test to run all of the above targets.
  • example to build and run the C++ example.
  • all to run everything.

Results are saved in the target directory.


Build the unit tests by entering this make command:

make test


The simple C++ example in examples/training.cpp uses the JSON for Modern C++ library class.

Build and run the example by entering this make command:

make example

This example adds the manifest tests/fixtures/training.json to the image file tests/fixtures/A.jpg using the sample private key and certificate in the tests/fixtures directory.

The example displays some text to standard out that summarizes whether AI training is allowed based on the specified manifest and then saves the resulting image file with attached manifest to target/example/training.jpg.

Supported file formats

ExtensionsMIME type
avivideo/msvideo, video/avi, application-msvideo
jpg, jpegimage/jpeg
mp4video/mp4, application/mp4 *
pdfapplication/pdf **
* Fragmented mp4 is not yet supported.** Read only


This package is distributed under the terms of both the MIT license and the Apache License (Version 2.0).

Note that some components and dependent crates are licensed under different terms; please check the license terms for each crate and component for details.

Contributions and feedback

We welcome contributions to this project. For information on contributing, providing feedback, and about ongoing work, see Contributing.