TransformerKt is a Kotlin coroutine wrapper library around media3.transformer:

Transformer is an API for editing media, including converting between formats (transcoding), applying changes like trimming a clip from a longer video, cropping a portion of the video frame, applying custom effects, and other editing operations

You can view the TransformerKt KDocs at docs.transformerkt.dev

• Using media3.transformer version 1.4.0-alpha01

Table of Contents

• Motivation
• Getting Started
• Usage
• Applying Effects
• Composition
• Demo App
• License

Motivation

The media3.transformer API is Java based and therefore relies on callbacks to notify the caller of the result of an operation. This library wraps the API in a Kotlin coroutine based API to make it easier to use. It exposes the Transformer API as either a suspend function or a Flow.

This library also includes some helpful extension functions to make it easier to use the API. See Usage for more information.

Note: Due to the way Transformer works, the coroutines must be launched on the Dispatchers.Main thread, otherwise the API will throw an IllegalStateException. Since it relies on the current thread to contain a Looper. While it is launched on the main-thread, Transformer delegates all the heavy lifting off of the main thread. See the docs for more information.

Getting Started

Add the dependency to your app level build.gradle.kts file:

dependencies {
    implementation("dev.transformerkt:transformerkt:3.6.0-beta01")
}

Or using Version Catalogs:

[versions]
transformerkt = "3.6.0-beta01"

[libraries]
transformerkt = { group = "dev.transformerkt", name = "transformerkt", version.ref = "transformerkt" }

Usage

First you should familiarize yourself with the Transformer Docs.

Inputs

Then you need an input video or image file. TransformerKt supports the following inputs:

• MediaItem.
• EditedMediaItem.
• A Uri pointing to somewhere on the device.
• A File object pointing to a file in the app's sand-boxed storage.
  • Warning: Getting a File object to a file outside of the app's storage will probably cause a permission error.

Now that you have your input sorted, there are two ways to consume this library.

Extension functions

A few extension functions have been added to the Transformer instance.

• suspend fun Transformer.start(): TransformerStatus.Finished
• fun Transformer.start(): Flow<TransformerStatus>

There are overloads for each of the supported inputs. For example:

suspend fun transform(context: Context, input: Uri) {
    val output = File(context.filesDir, "output.mp4")
    val transformer = TransformerKt.build(context) {
        setVideoMimeType(MimeTypes.VIDEO_H264)
    }
    val result = transformer.start(input, output) { progress ->
        // Update UI progress
    }
    when (result) {
        is TransformerStatus.Failure -> TODO()
        is TransformerStatus.Success -> TODO()
    }
}

Or you can use the Flow version instead:

fun transform(context: Context, input: Uri) {
    val output = File(context.filesDir, "output.mp4")
    val transformer = Transformer.build(context) { setVideoMimeType(MimeTypes.VIDEO_H264) }
    transformer.start(input, output).collect { status ->
        when (status) {
            is TransformerStatus.Progress -> TODO()
            is TransformerStatus.Success -> TODO()
            is TransformerStatus.Failure -> TODO()
        }
    }
}

Applying Effects

Starting with Media3 1.1.0-alpha01 the Transformer library changed the way you apply effects. Instead of applying the effects to the Transformer.Builder you now create a EditedMediaItem and apply the affects there.

To make that API a bit easier, an extension function .edited {} has been added to MediaItem.Builder:

val editedMediaItem = MediaItem.Builder()
    .setUri(Uri.parse("https://example.com/video.mp4"))
    .setMediaId("Foo")
    .edited {
        setRemoveAudio(true)
    }

val result = TransformerKt.build(context).start(editedMediaItem, File("output.mp4"))

Or directly from a [MediaItem] instance:

val editedMediaItem = MediaItem
    .fromUri(Uri.parse("https://example.com/video.mp4"))
    .edited {
        setRemoveAudio(true)
    }

val result = TransformerKt.build(context).start(editedMediaItem, File("output.mp4"))

Composition

Transformer now supports Composition which allows you to combine multiple inputs into a single output. You can apply effects to the whole composition or on a per input basis:

data class MyComplexItem(val tag: String, val uri: Uri, val startMs: Long, val endMs: Long)

val items: List<Uri>
val complexItems: List<MyComplexItem>
val endCredits: File
val audioOverlay: File
val composition = compositionOf {
    // Apply effects to the whole composition
    effects {
        resolution(1920, 1080, LayoutScale.Fit)
    }

    // Create a sequence of inputs
    sequenceOf {
        items(items) { uri ->
            effects {
                bitmapOverlay(context, R.drawable.watermark) {
                    setScale(.2f, .2f)
                    setOverlayFrameAnchor(.8f, .8f)
                }
            }
        }

        items(
            items = complexItems,
            selector = { it.uri },
            configure = { complexItem ->
                // Configure the MediaItem instance
                setTag(complexItem.tag)
                setClippingConfiguration(complexItem.startMs, complexItem.endMs)
            },
        ) { complexItem ->
            setRemoveAudio(true)

            effects {
                speed(2f)
                brightness(0.5f)
            }
        }

        item(endCredits)
    }

    sequenceOf(isLooping = true) {
        item(audioOverlay)
    }
}

Checkout the TransformerRepo.kt file for more examples.

Included Effects

Currently only one custom effect is included in this library. It is the AudioProcessor called VolumeChangeProcessor. It allows you to change the volume of the audio track in the video.

val volumeChange = volumeChangeEffect(inputChannels = 2, volume = 0.5f)
// ... Add to your EditedMediaItemSequence

If you provide a VolumeChangeProcessor you are able to customize the volume based on the elapsed time in the output video:

val volumeChange = volumeChangeEffect(inputChannels = 2) { elapsedMs ->
    when {
        elapsedMs < 1000 -> 0f
        elapsedMs < 2000 -> 0.5f
        else -> 1f
    }
}

For convenience there is a fadeAudioOutEffect that will fade the audio out over a given duration:

val volumeChange = fadeAudioOutEffect(
    totalDurationUs = 10_000_000,
    inputChannels = 2,
    initialVolume = 0.8f,
    finalVolume = 0.3f,
    fadeDurationUs = 2_000_000,
)

This effect will fade the audio from 0.8f to 0.3f over the course of 2_000_000 microseconds (2 seconds) from the end of the video.

Demo App

A demo app is included in the demo module. It is a simple app that allows you to select a HDR video and convert it do a SDR video.

To run the demo app you can follow these steps:

git clone git@github.com:jordond/transformerkt.git transformerkt
cd transformerkt
./gradlew assembleRelease

Then install the demo/build/outputs/apk/release/demo-release.apk file on your device.

License

See LICENSE