Re: too mach distortion on color gradient
Alexey Fisher <[email protected]> Fri, 08 Oct 2010 14:02:05 +0200
| Newsgroups | gmane.comp.multimedia.ogg.theora.devel |
|---|---|
| Message-ID | <1286539325.5264.80.camel@zwerg> |
Hallo Monty, Am Donnerstag, den 07.10.2010, 17:04 -0400 schrieb [email protected]: > First off, Ptalar does handle these cases better than Thusnelda did. > Second, the whole problem stems from the lowest AC coeffs not getting > enough bits, and I'd expect that to be a result of coarse quantizers. > What were the exact command lines used to encode? I'm only used to > seeing this effect in fairly low bitrate encodes. i'm not in sintel team, so i can't answer this question. I send an email Ton Roosendaal (ton blender.org), will see if we will get some response. I still think it will be good (especially for advocacy) to make libtheora writhe some encoding options to the header. Currently it write only encoder version. Relevant info is bitrate/quality and passes. So it will look like: k2000p2 (2000 kbit, 2 pass) or q60p2 (quality 60, 2 pass) every thing together will be: Xiph.Org libtheora 1.2.0alpha 20100924 (Ptalarbvorm) q60p2 ^^^^^ Alexey