DefaultVideoFrameProcessor

const val WORKING_COLOR_SPACE_DEFAULT = 0: Int

Use BT709 color primaries with the standard SDR transfer function (SMPTE 170m) as the working color space.

Any SDR content in a different color space will be transferred to this one.

const val WORKING_COLOR_SPACE_LINEAR = 2: Int

The working color space will have the same primaries as the input and a linear transfer function.

This option is not recommended for SDR content since it may lead to color banding since 8-bit colors are used in SDR processing. It may also cause effects that modify a frame's output colors (for example overlays) to have incorrect output colors.

const val WORKING_COLOR_SPACE_ORIGINAL = 1: Int

Use the original color space of the input as the working color space when the input is SDR.

Tonemapped HDR content will be represented with BT709 color primaries and the standard SDR transfer function (SMPTE 170m).

No color transfers will be applied when the input is SDR.

fun flush(): Unit

Flushes the VideoFrameProcessor}.

All the frames that are #registerInputFrame() registered prior to calling this method are no longer considered to be registered when this method returns.

methods invoked prior to calling this method should be ignored.

The downstream frame consumer must be flushed before this instance is flushed, and stop accepting input until this DefaultVideoFrameProcessor instance finishes flushing.

After this method is called, any object consuming texture output must not access any output textures that were rendered before calling this method.

fun getInputSurface(): Surface!

Returns the input Surface, where VideoFrameProcessor consumes input frames from.

The frames arriving on the Surface will not be consumed by the VideoFrameProcessor until registerInputStream is called with INPUT_TYPE_SURFACE.

For streams with INPUT_TYPE_SURFACE, the returned surface is ready to use immediately and will not have a default buffer size set on it. This is suitable for configuring a android.media.MediaCodec decoder.

For streams with INPUT_TYPE_SURFACE_AUTOMATIC_FRAME_REGISTRATION, set a listener for the surface becoming ready via setOnInputSurfaceReadyListener and wait for the event before using the returned surface. This is suitable for use with non-decoder producers like media projection.

fun getPendingInputFrameCount(): Int

Returns the number of input frames that have been made available to the VideoFrameProcessor but have not been processed yet.

@VisibleForTesting
fun getTaskExecutor(): VideoFrameProcessingTaskExecutor!

Returns the task executor that runs video frame processing tasks.

fun queueInputBitmap(
    inputBitmap: Bitmap!,
    timestampIterator: TimestampIterator!
): Boolean

Provides an input to the .

Can be called many times after #registerInputStream registering the input stream to put multiple frames in the same input stream.

An implementation could Bitmap#recycle the passed in , so it may not be suitable for reuse.

The passed in Bitmap argument cannot be reused after this method returns true, as it will be recycled by the processing pipeline.

fun queueInputTexture(textureId: Int, presentationTimeUs: Long): Boolean

Provides an input texture ID to the VideoFrameProcessor.

It must be only called after setOnInputFrameProcessedListener and registerInputStream have been called.

fun redraw(): Unit

Updates an Listener#onOutputFrameAvailableForRendering available frame with the modified effects.

This method can be called from any thread.

DefaultVideoFrameProcessor keeps track of the redraw requests received. If a call to redraw is made when another redraw request is ongoing, the new request will be performed later when the ongoing redraw completes, and this method will return immediately.

fun registerInputFrame(): Boolean

Informs the VideoFrameProcessor that a frame will be queued to its input surface.

Must be called before rendering a frame to the input surface. The caller must not render frames to the input surface when false is returned.

fun registerInputStream(
    @VideoFrameProcessor.InputType inputType: Int,
    format: Format!,
    effects: (Mutable)List<Effect!>!,
    offsetToAddUs: Long
): Unit

Informs the VideoFrameProcessor} that a new input stream will be queued with the list of Effects to apply to the new input stream.

After registering the first input stream, this method must only be called after the last frame of the already-registered input stream has been #registerInputFrame registered, last bitmap #queueInputBitmap queued or last texture id * #queueInputTexture queued.

This method blocks the calling thread until the previous calls to this method finish, that is when Listener#onInputStreamRegistered(int, Format, List) is called after the underlying processing pipeline has been adapted to the registered input stream.

Using HDR colorInfo requires OpenGL ES 3.0 and the EXT_YUV_target OpenGL extension.

Effects are applied on COLOR_RANGE_FULL colors with nullhdrStaticInfo.

If either colorInfo or outputColorInfoisTransferHdr are HDR}, textures will use GL_RGBA16F and GL_HALF_FLOAT. Otherwise, textures will use GL_RGBA and GL_UNSIGNED_BYTE.

If input coloris HDR, but outputColorInfo is SDR, then HDR to SDR tone-mapping is applied, and outputColorInfo's colorTransfer must be COLOR_TRANSFER_GAMMA_2_2 or COLOR_TRANSFER_SDR. In this case, the actual output transfer function will be in COLOR_TRANSFER_GAMMA_2_2, for consistency with other tone-mapping and color behavior in the Android ecosystem (for example, MediaFormat's COLOR_TRANSFER_SDR_VIDEO is defined as SMPTE 170M, but most OEMs process it as Gamma 2.2).

fun release(): Unit

Releases all resources.

If the VideoFrameProcessor is released before it has ended, it will attempt to cancel processing any input frames that have already become available. Input frames that become available after release are ignored.

This method blocks until all resources are released or releasing times out.

This VideoFrameProcessor instance must not be used after this method is called.

fun renderOutputFrame(renderTimeNs: Long): Unit

Renders the oldest unrendered output frame that has become * Listener#onOutputFrameAvailableForRendering(long, boolean) available for rendering at the given renderTimeNs}.

This will either render the output frame to the #setOutputSurfaceInfo output surface, or drop the frame, per renderTimeNs}.

This method must only be called if renderFramesAutomatically} was set to *false using the and should be called exactly once for each frame that becomes Listener#onOutputFrameAvailableForRendering(long, boolean) available for rendering.

The renderTimeNs} may be passed to EGLExt#eglPresentationTimeANDROID depending on the implementation.

If texture output is set, calling this method will be a no-op.

fun setInputDefaultBufferSize(width: Int, height: Int): Unit

Sets the default size for input buffers, for the case where the producer providing input does not override the buffer size.

When input comes from a media codec it's not necessary to call this method because the codec (producer) sets the buffer size automatically. For the case where input comes from CameraX, call this method after instantiation to ensure that buffers are handled at full resolution. See setDefaultBufferSize for more information.

This method must only be called when the VideoFrameProcessor is created with INPUT_TYPE_SURFACE.

fun setOnInputFrameProcessedListener(
    listener: OnInputFrameProcessedListener!
): Unit

Sets the OnInputFrameProcessedListener.

fun setOnInputSurfaceReadyListener(listener: Runnable!): Unit

Sets a listener that's called when the input surface is ready to use.

fun setOutputSurfaceInfo(outputSurfaceInfo: SurfaceInfo?): Unit

Sets the output surface and supporting information. When output frames are rendered and not dropped, they will be rendered to this output .

The new output is applied from the next output frame rendered onwards. If the output is null}, the VideoFrameProcessor} will stop rendering pending frames and resume rendering once a non-null is set.

If the dimensions given in do not match the * Listener#onOutputSizeChanged(int,int) output size after applying the final effect the frames are resized before rendering to the surface and letter/pillar-boxing is applied.

The caller is responsible for tracking the lifecycle of the SurfaceInfo#surface including calling this method with a new surface if it is destroyed. When this method returns, the previous output surface is no longer being used and can safely be released by the caller.

If texture output is set, calling this method will be a no-op.

fun signalEndOfInput(): Unit

Informs the VideoFrameProcessor that no further input frames should be accepted.