Fabio
2008-Jan-24 01:50 UTC
[theora] Re: [dyne:bolic] Concerning the (correct) use of Theora in FreeJ
> Neither Theora nor h.264 have yet > reached their practical efficiency limits, and we should expect > continuing improvement from both for several more years.On your document here: http://web.mit.edu/xiphmont/Public/theora/demo.html you show the result of the improved rate managemet in the theora codec. Can you tell which are the encoders that produced the two video matrix-580.ogg and matrix-55.ogg? I thought the first was the codec from theora trunk, and the second that of an old revision of theora-thusnelda. Also, is the original .y4m of the matrix movie available for download for testing? I tried some revision of theora-thusnelda encoder (including the last one), but seems that its quality is still lower than current theora trunk codec.
xiphmont@xiph.org
2008-Jan-24 13:57 UTC
[theora] Re: [dyne:bolic] Concerning the (correct) use of Theora in FreeJ
On Jan 24, 2008 4:50 AM, Fabio <fabio.ped@libero.it> wrote:> you show the result of the improved rate managemet in the theora codec. Can you tell which are the encoders that produced the two video matrix-580.ogg and matrix-55.ogg?Matrix 55 and 580 were both slightly modified mainline running with some canned settings. 'Canned' in the sense that I rigged the demo by using internal settings that I knew would work well for the Matrix clip, not code that knew enough to make those settings automatically. That, of course, is the real work. None of the demo clips came from Thusnelda.> Also, is the original .y4m of the matrix movie available for download for testing?No, for obvious reasons. I'll tell you it's a straight rip of chapter 28 of the matrix DVD de-letterboxed and scaled to 640x272 (to be identical with the test clips DOOM9 likes to use).> I tried some revision of theora-thusnelda encoder (including the last one), but seems that its quality is still lower than current theora trunk codec.In SVN right now, thusnelda is half way to getting new mode selection code and is mostly choosing block coding modes randomly. It is not a stable branch, and will not even build most of the time. It's merely my own personal SVN sandbox for now. When enough of the changes are stable, they'll go to mainline. Monty
Possibly Parallel Threads
- Re: [dyne:bolic] Concerning the (correct) use of Theora in FreeJ
- Theora file size
- [LLVMdev] multithreaded performance disaster with -fprofile-instr-generate (contention on profile counters)
- libtheora 1.1alpha1 (thusnelda) release
- libtheora 1.1alpha1 (thusnelda) release