Adapting Grafika RecordFBOActivity to work with Android GPUImage

Question

I have an application that is using the Android port of GPUImage as the OpenGL Renderer and manager of several filters.

It currently does not have a video implementation, so I am trying to adapt the RecordFBOActivity from the Google grafika repository to work with the GPUImage architecture.

The base GPUImage class manages the GLContext and GLSurfaceView, and the GPUImageRenderer class implements the Renderer class.

This is the class where I am trying to adapt the RenderThread from the RecordFBOActivity of grafika. There are a few problems.

First, in the preparegl() method, I am passing a SurfaceTexture instead of a Surface, as GPUImage doesn't use the SurfaceHolder at all (I think I can implement it, but am trying not to change the base code too much, as i would like to push back my implementation to the aforementioned repo). I know that WindowSurface.java has an overloaded method to construct a WindowSurface from a SurfaceTexture as well as a Surface, but if I do this the mSurface iVar is always null, as I never have a surface to pass to it, which causes a NPE in the makeCurrent() method of recording.

Second, GPUImage attaches itself to a GLSurfaceView, not a SurfaceView like the grafika example uses, so I'm a little uncertain if there are any low level inconsistencies that may be causing conflicts for me...

Third, and I think this is the main issue, at least at the moment, is that I can't seem to reconcile the camera preview of GPUImage with the WindowSurface of grafika. If I comment out the prepareGl() method, the setUpSurfaceTexture() of GPUImage sets the preview texture of the camera from the SurfaceTexture that is created by glGenTextures() and the preview works fine.. as well as being attached to the filter render chain. However, if I try to call the prepareGL() method, and pass the exact same SurfaceTexture to the constructor of mWindowSurface, the camera service dies and i get a EGL_BAD_SURFACE error.

Long question, with a few moving parts, I know... Will attempt to edit/update as I can clarify issues and approaches to myself. But would love if anyone has any thoughts/interrogations... particularly @fadden :D

Answer 1

I was also trying to achieve the same thing and have tried what fadden has suggested. Tried to integrate CameraSurfaceRenderer functionality to GPUImageRenderer. The preview is fine but the recording is just a video with black frames. EGL14.eglGetCurrentContext() returns null for following call and my guess is if a new context is created it will not be same as what GPUImage might have

mVideoEncoder.startRecording(new TextureMovieEncoder.EncoderConfig(
                                    mOutputFile, 640, 480, 1000000, EGL14.eglGetCurrentContext()));

@Jesses.co.tt were you able to achieve it?

(as I can't add comment it is added as an answer).