How to use b-frames: 1 or 2?

[Teegedeck]

No, it won't. And the resulting filesize will be larger than that of an encode without b-frames. B-frames were meant to be compressed more than p-frames by design, and any attempt to use them in a different way will result in a net loss in quality.

Please, believe me that there isn't much room for optimization in the b-frame department; b-frames are fine at the default setting.

[dooWohcysP]

Understood.

...
[Running to stop the encoding of a 1000000 frames avs]
...


I won't mess up with XviD settings again! :o


I forgot:

THANKS! :)

[Sharktooth]

For fixed quant encodings just disable b-frames (if you dont care of the filesize).

[redfordxx]

Hi all,

after accepting, that the default Bframe settings are mostly the best possible...

...when still one wants to play a little with it (maybe to get a feeling how it works):


Do I suppose correctly that B-f Q should be *always* the same or (better?) higher than *both* surrounding non-B-f? Otherwise the benefit is lost when trying to make better frame (B) from worse one (P), which requires much more texture data?

Another question (in fact, same valid for Pframes):
When I have serie of frametypes(with quantizers) like
P(Q2) B(Q4) P(Q3)
and the B can be pretty well reconstructed from P-f's...
Is the resulting quality of the B more as if used Q3?

[Ark]

B-frames Q *should* be higher than I and P frames Q, that's the expected use of them, expecially because they usually are superior in number to all P-I frames (but that's not always true, it entirely depends on the source and encoder/user decisions), so the better-quality-at-same-filesize effect they give would be lost lowering their Q too much.

For the second question, sure if a b-frame is surrounded by higher qulity P-frames, the results is better than with lower quality p-frames (Q4,5...), though i don't think you can gain a whole Q quality-wise, but that's only my thought!