Support YUV CVPixelBuffer output about metalpetal HOT 6 CLOSED

Will consider the "render to YUV" feature.

from metalpetal.

Output to CVPixelBuffer at the end of the chain, would be great. So we can use that output for other stuff like making videos or publishing to a live stream or something else.

from metalpetal.

MTIContext object supports rendering a MTIImage to a CVPixelBuffer. However it can only render to a BGRA pixel buffer for now.

from metalpetal.

i see CoreImage support render to yuv CVPixelBuffer. but it cost too much cpu. more than 100% cpu will be used.

  [ciContext render:outimage toCVPixelBuffer:outPixelBuffer];

is there any way to reduce the cpu use?

from metalpetal.

it seems it is because of https://github.com/YuAo/YUCIHighPassSkinSmoothing, move the discussion to YuAo/YUCIHighPassSkinSmoothing#11

from metalpetal.

@omarojo @notedit Render to YUV pixel buffer is implemented on supported hardware (MTLFeatureSet_iOS_GPUFamily3_v1). ref: 2226c0f

Not fully tested, use with caution.

Here's an example:

NSError *error;
CVPixelBufferRef pixelBuffer;
CVPixelBufferCreate(kCFAllocatorDefault, outputImage.size.width, outputImage.size.height, kCVPixelFormatType_420YpCbCr8BiPlanarFullRange, (__bridge CFDictionaryRef)(@{(id)kCVPixelBufferIOSurfacePropertiesKey: @{}}), &pixelBuffer);
[context renderImage:outputImage toCVPixelBuffer:pixelBuffer error:&error];
//use the pixelBuffer
//....
//release the pixelBuffer
CVPixelBufferRelease(pixelBuffer);

from metalpetal.