Skip to main content

Creating Frame Processor Plugins (C++)

C++ Frame Processor Plugins

C++ based Frame Processor Plugins are built with JSI and can be directly called from a Frame Processor as native JSI functions can be shared across contexts. The benefits of C++ based Frame Processor Plugins lie in the nature of C++ based code:

  • Better Performance due to less marshalling required and lower level access
  • Shared platform code - e.g. OpenCV Frame Processor Plugins only need to be implemented once

For example:

// 1. Include headers
#include "frameprocessor/FrameHostObject.h"

// 2. Create C++ func
auto myPlugin = [=](jsi::Runtime& runtime,
const jsi::Value& thisArg,
const jsi::Value* args,
size_t count) -> jsi::Value {
auto frame = args[0].asObject(runtime).asHostObject<FrameHostObject>(runtime);
// Unwrap the Frame, then do your Frame Processing here, and return any JSI value as a result.
// For example, you can run very efficient OpenCV tasks here.
return jsi::Value(42);
};
// 3. Wrap C++ func in jsi::Function
auto jsiFunc = jsi::Function::createFromHostFunction(runtime,
jsi::PropNameID::forUtf8(runtime,
"myCppPlugin"),
1,
myPlugin);
// 4. Add it to global so it can be called from JS
runtime.global().setProperty(runtime, "myCppPlugin", jsiFunc);

Then to call that function:

const frameProcessor = useFrameProcessor((frame) => {
'worklet'
const result = global.myCppPlugin(frame)
console.log(`C++ result: ${result}`) // <-- 42
}, [])

To include VisionCamera's C++ library in your plugin's C++ code, you need to link it and include the headers.

VisionCamera is built using CocoaPods, which allow you to include local dependencies. To add VisionCamera to your package's .podspec, just add the dependency:

s.dependency "VisionCamera"

Then include it in your C++ code:

#include "Frame Processor/FrameHostObject.h"

🚀 Next section: Browse Community Plugins