When used with ffmpeg leverage vaapi for much faster generation

[zkvsky]

I've been experimenting with hw accelerated ffmpeg thumbnailer via changing ffmpeg command to:

    local ffmpeg_command = {
        "ffmpeg",
        "-loglevel", "quiet",
        "-noaccurate_seek",
        "-hwaccel", "vaapi",
        "-hwaccel_device", "/dev/dri/renderD128",
        "-hwaccel_output_format", "vaapi", 
        "-ss", format_time(timestamp, ":"),
        "-i", file_path,

        "-frames:v", "1",
        "-an",

        "-vf", ("scale_vaapi=%d:%d,scale_vaapi=format=rgb32,hwdownload,format=rgb32"):format(size.w, size.h),
        "-c:v", "rawvideo",
        "-pix_fmt", "bgra",
        "-f", "rawvideo",

        "-y", output_path
    }```


And it is much much faster. could it be used when ffmpeg is picked?

[Steven79203]

I tested these same configurations but didn't worked properly without modifications. Here's my config if anyone with similar specs want to try: CPU: 3th Gen Intel Ivybridge, GPU: Intel Ivybridge (Gen 7) Integrated Graphics

local ffmpeg_command = {
        "ffmpeg",
        --"-loglevel", "quiet",
        "-noaccurate_seek",
        "-ss", format_time(timestamp, ":"),
        "-hwaccel_output_format","vaapi",
        "-hwaccel_device","/dev/dri/renderD128",
        "-hwaccel", "vaapi",
        "-i", file_path,
        "-frames:v", "1",
        "-an",


        "-vf", ("scale_vaapi=%d:%d,scale_vaapi=format=rgb32,hwdownload,format=nv12"):format(size.w, size.h),
        "-c:v", "rawvideo",
        "-pix_fmt", "bgra",
        "-f", "rawvideo",

        "-y", output_path
}

Works like a charm.