scaling and pillarboxing the input from the grabber

gistfile1.txt

	ffmpeg \
	  -y \
	  -v debug \
	  -report \
	  -nostdin \
	  -init_hw_device "vaapi=intel:/dev/dri/renderD128" \
	  -hwaccel vaapi \
	  -hwaccel_output_format vaapi \
	  -hwaccel_device intel \
	  -filter_hw_device intel \
	  -probesize 10M \
	  -analyzeduration 10M \
	  -f v4l2 \
	  -video_size 1280x1024 \
	  -framerate 30 \
	  -i /dev/video2 \
	  -itsoffset 0.064 \
	  -f alsa -sample_rate 48000 \
	  -channels 2 \
	  -i hw:1 \
	  -threads:0 0 \
	  -aspect 16:9 \
	  -filter_complex "[1:a]channelsplit=channel_layout=stereo[left][right];[0:v]scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:-1:-1:color=black,format=nv12,hwupload [vout]"\
	  -map "[vout]" \
	  -c:v:0 h264_vaapi \
	  -rc_mode CBR \
	  -g 30 \
	  -x264opts "keyint=30:min-keyint=30:no-scenecut" \
	  -maxrate:v:0 5000k \
	  -bufsize:v:0 8192k \
	  -b:v:0 3000k \
	  -qmin:v:0 1 \
	  -map "[left]:1" \
	  -ac 1 \
	  -strict \
	  -2 \
	  -c:a aac \
	  -b:a 128k \
	  -ar 48000 \
	  -map "[right]:2" \
	  -ac 1 \
	  -strict \
	  -2 \
	  -c:a aac \
	  -b:a 128k \
	  -ar 48000 \
	  -y \
	  -f mpegts /tmp/tmp.ts

I've tested the above pipeline with the MS2130 and a 12801024, resp. 1024768 source resolution.

Note that the grabber's supported output resolutions apparently are the following according to the YiuzukiHD github docs:

• 1920*1080 (16/90)
• 1600*1200 (4/3)
• 1360*768 (16/9-ish)
• 1280*1024 (5/4)
• 1280*960 (4/3)
• 1280*720 (16/9)
• 1024*768 (4/3)
• 800*600 (4/3)
• 720*576 (5/4)
• 720*480 (3/2)
• 640*480 (4/3)

All of these should be supported at 10, 20, 30, 50 and 60 Hz.

At first sight, it looks as if we can ignore resolution changes as long as the source aspect ratio stays the same. That would be wonderful, considering most projectors will negotiate the same aspect ratio over and over again

This kind of pseudocode should work then:

previous_signal
current_signal
previous_source_resolution
current_source_resolution
current_source_aspect_ratio
previous_source_aspect_ratio

• if ( current_signal && 1.25 <= current_source_aspect_ratio <=1.78 )
  • if ( !previous_signal | current_source_aspect_ratio != previous_source_aspect_ratio)
    • restart ffmpeg pipeline; parameter: current_source_resolution
  • else
    • do nothing
• else
  • generic FOSDEM slide on stream