I've been making huge strides with libavpp getting more of the FFmpeg C API examples ported over. The fact that the hardware transcoding examples are working now is a huge leap in progress, which means we can bring it back to audioviz and potentially stop relying on popen()-ing ffmpeg which is definitely a performance bottleneck.
The fastest encode speed I could get from audioviz was about 6x using the NVIDIA GPU to render (prime-run) and the h264_vaapi encoder. The audioviz visualizer was just a bare StereoSpectrum, the same one in spectrum-test.cpp. Linking to FFmpeg libs can potentially bring even more performance!
I've been making huge strides with libavpp getting more of the FFmpeg C API examples ported over. The fact that the hardware transcoding examples are working now is a huge leap in progress, which means we can bring it back to audioviz and potentially stop relying on
popen()-ingffmpegwhich is definitely a performance bottleneck.The fastest encode speed I could get from audioviz was about 6x using the NVIDIA GPU to render (
prime-run) and theh264_vaapiencoder. The audioviz visualizer was just a bareStereoSpectrum, the same one inspectrum-test.cpp. Linking to FFmpeg libs can potentially bring even more performance!