The idea is very simple: get an interpolated frame from given frames. For example, given the current frame and next decoded frame, using linear interpolation, we can get a frame at any time between curent and next decoded frame pts. At least 2 FBO is required. Maybe we can use current and previous decoded frame.

No FBO is possible too, but more sampler2D uniforms are required. Maybe we can use User Configurable Shaders.

There are some special formats used by some vendors, tiled layout for example. Tiled rendering can benefits from those layouts. But we should convert it to the standard layouts for opengl rendering. If we can get plane textures from OMX (0-copy), it's very simple to convert tiled to standard using GLSL. Again we can use User Configurable Shaders technology, creating a sampling function for tiled formats. The tiled info can be stored in VideoFrame.mataData.