On Jan 17, 2013, at 11:52, Harald Jordan wrote: > first problem for me. The input for an encoder is anyway always a decoded > frame, so it always has the “perfect” quality. So assumed we always want to > have the best available quality, basically the assumption that the qscale > settings depend on your input is not correct, right?
Not exactly, I think - at least not for input that is decoded from a lossy compression method. It is my experience (using existing software, not tweaking qmin/qmax/qscale) that such re-encoding such frames with too high a quality setting can be counterproductive. I think of it as exaggerating the (de)compression artefacts that can result in a file size increase rather than decrease. > Now, when I write an encoder using libavcodec that is always fed from yuv422p > pictures, what sense do those settings make? Control the level of detail in (and the size of) your encoded content, I presume? R _______________________________________________ Libav-user mailing list [email protected] http://ffmpeg.org/mailman/listinfo/libav-user
