[MPlayer-users] [BUG] ratecontrol.c:693: ff_rate_estimate_qscale: Assertion `pict_type == rce->new_pict_type' failed.

[Corey Hickey]

vitko wrote:
>> It could be a bug, but I would just advise you to set vb_strategy to
>> whatever you want on 1st pass, and not set it in second pass.
> 
> You are right and I haven't done my homework right, man mencoder says
> 
> ...
> vb_strategy=<0-2> (pass one only)
>                    ^^^^^^^^^^^^^
> ...
> 
> Maybe it should say "never ever use vb_strategy in pass 2"?

I was thinking a while ago that it would be better to simply ignore 
vb_strategy on later passes (and maybe print a warning). I'll propose 
that on ffmpeg-devel once CVS or SVN is working again.

> When I delete vb_strategy from pass 2 lavc options, all goes smooth and 
> fine.

By the way, you should try out vb_strategy=2 and/or turbo for the first 
pass. Using turbo will speed up your first pass very much and have a 
nearly immeasurable effect on quality. vb_strategy=2 will improve 
quality somewhat (much better than vb_strategy=1) but will slow the 
first pass down a lot. Using both options together will result in the 
first pass encoding somewhat more quickly and the second pass looking 
somewhat better than if you hadn't used either.

Yeah, I know that's not in the docs. I've been lazy about them.

Also, if you want to try something else, tell me whether you like this 
option.
vrc_eq=(tex+10^8*mcVar)^0.6

You may have to put quotes around it so your shell doesn't try to 
interpret the parentheses.

Use it in both passes. You should see a large improvement in visual 
quality in high-motion scenes without a significant decrease in quality 
in low-motion scenes.

-Corey