The Raspberry Pi will do hardware accelerated h264 encoding when recording video from the camera board. How can I encode h264 (using hardware acceleration) when the source images are not coming directly from the camera?
Are there any command line tools that support this?
If no, what APIs should I be looking at to make this possible?
GStreamer is included in Raspbian and with its OpenMAX plugin it will use the hardware encoding capabilities of the Raspberry Pi.
See this link for a tutorial on doing what you're looking for: https://www.raspberrypi.org/forums/viewtopic.php?t=72435
If you're interested in transcoding, I've just posted an answer to another question that might interest you: What speed can I expect from the hardware-H264-encoding?
Looks like by compiling ffmpeg by yourself, on RPi3B+ you can get
ENCODING HIGH-COMPLEXITY 30 FPS VIDEO FROM A 1920 X 1080 JPEG IMAGE SEQUENCE – SIGNIFICANT MOTION AND INTRAFRAME DETAIL – HIGH QUALITY SETTINGS:
Software-Based H.264 Encoding (CPU): 2.6 FPS (11.5 times slower than real time)
Hardware-Based H.264 Encoding (GPU): 6.3 FPS (4.8 times slower than real time)
ENCODING HIGH-COMPLEXITY 30 FPS VIDEO FROM A 640 X 480 JPEG IMAGE SEQUENCE – SIGNIFICANT MOTION AND INTRAFRAME DETAIL – HIGH QUALITY SETTINGS:
Software-Based H.264 Encoding (CPU): 18 FPS (1.7 times slower than real time)
Hardware-Based H.264 Encoding (GPU): 38 FPS (1.3 times FASTER than real time)
The script to help achieve all this is in a forum thread.
raspiyuv -o - | rpi-encode-yuv(roughly) where therpi-encode-yuvtool is from here. The frame size is hard-coded in this tool, so you need to change it at the beginning of the C file and recompile. Unfortunately I haven't managed to get a properly aligned output yet, asraspiyuvdoesn't seem to encode the frame size into its output. I get a gradually shifting picture that eventually turns to garbage. Still working on it. – Szabolcs Jun 6 '14 at 3:20