I have been transcoding a group of panoramic videos using ffmpeg2theora 0.26
i just completed transcoding Naples Cycle HDV.vwmovie
the resulting video has some strange colors in it
the source video can be got from http://www.eyesee360.com/videowarp/examples/
the resulting images and the theora file can be download from
http://www.lightningridgecommunity.com/images/
tom at tom-laptop:~/Videos/krpano$ ffmpeg2theora -V1000k -A64k "Naples
Cycle HDV.vwmovie"
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'Naples Cycle HDV.vwmovie':
Metadata:
major_brand : qt
minor_version : 537199360
compatible_brands: qt
Duration: 00:02:00.90, start: 0.000000, bitrate: 8917 kb/s
Stream #0.0(eng): Audio: aac, 44100 Hz, stereo, s16, 96 kb/s
Stream #0.1(eng): Video: h264, yuv420p, 1554x590, 8808 kb/s, 25 fps, 25 tbr,
600 tbn, 1200 tbc
0:02:00.90 audio: 65kbps video: 1000kbps, time elapsed: 00:23:17
tom_a_sparks
On Sun, Feb 14, 2010 at 12:49 PM, Tom Sparks <tom_a_sparks at yahoo.com.au> wrote:> I have been transcoding a group of panoramic videos using ffmpeg2theora 0.26 > i just completed transcoding Naples Cycle HDV.vwmovie > the resulting video has some strange colors in it > > the source video can be got from http://www.eyesee360.com/videowarp/examples/ > the resulting images and the theora file can be download from http://www.lightningridgecommunity.com/images/ > > tom at tom-laptop:~/Videos/krpano$ ffmpeg2theora -V1000k -A64k "Naples Cycle HDV.vwmovie" > Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'Naples Cycle HDV.vwmovie':Hi Tom, I don't think it is a _bug_, I think its just poor quality. You're encoding 1554*590 4:2:0 @25fps into 1-megabit, which is 0.04 bits per pixel. Nothing is going to look great at that low a rate, especially high motion with a strictly buffer limited stream. Your H.264 input is at ~9megabits and even that doesn't look all that perfect. I think all you're seeing in this case is that the encoder is running out of bits so thoroughly that it is making the color information very blocky. This chunkyness produces the somewhat off colors. The current development encoder is more intelligent about bit allocation decisions in smooth areas. When I combine the development encoder with a two-pass encode I get this: http://myrandomnode.dyndns.org:8080/~gmaxwell/theora/Naples_Cycle_ptalarbvorm_twopass.ogv and I think that looks a lot better.