I am in the process of encoding some PAL SVHS- and HI8-tapes to divx5. To my surprise the quality is very good, especially the hi8-tapes which nearly meet dvd-standards. Most stuff are old personal videos and there are also some movies and cartoons/anime.
I still have one problem: If recording anime, I get lots of ghost-shadows. Thats not a telesyn-problem, because telesyn and 25fps-PAL exclude each other. And it is not a interlaced-problem because this problem persists even when I drop field 2 and therefore no longer are using interlace. When I look closer, it seems to be an effect encoded right into the source-material at fullfame-basis - does that make sense?
My personal howto, which proved usefull in some local forums:
While googling the web I didn´t came across howtos or guides about this and tried my own ideas. But if you have some comments or places to look for, please participate in this thread.
You need Virtualdub, huffyuv, divx5 and some mp3-Codec. You might want to look into the Nimo-Codec-Pack, it brings many usefull stuff in one packet. Get Donald Grafts smart smoother
My workflow looks like this:
record using Virtualdub and huffyuv lossless compression.
cropping from 704x576 down to 688x550 to
-remove distortion at the borders
-keep a PAL-aspectratio of 5:4
-keep AVI-limits, xres 16bit-boundry, y-red at 2bit-boundry
this produces around 400MB/min at perfect quality. I wont explain segmented AVI, the 2GB or 4GB-limit, thats beyond the scope of this faq, ask your favourite searchengine.
I prefer a NTSC-ratio of 4:3 because thats the aspectratio of computermonitors and as divx is still a computeronly-solution. I want something which works outofthebox on a computer and this means a resolution of 688x516, 512x384, 352x264 of 320x240.
Always set audio and video to full process.
1-pass-quality-based at 60% to 80% gives good quality, but no control about the final size.
2-pass is harder and slower to use, but gives fine control about the final size.
Using "Psychovisual Enhancement Strong" seems to utilize the bitrate more efficient, I would say 3-10%. Thats work in progress, you might get a bloody nose in the future, but I still go for it.
to make an AVI at 688x516 or 512x384:
-audiocompression: mp3, 44.100khz, 128kbit stereo - videotapes can have an excellent audioquality
-videofilter: deinterlace mode blend
-videofilter: resize to final resolution, bicubic
-videofilter: Donald Grafts smart-smoother, diam 7, str 20, inter no
-videocompression: divx5, 1-pass-quality-based, 70%, Psycho-Enh Strong
-save to AVI, takes ~8x of playing time on a 1Ghz-system
to make an AVI at 352x264 or 320x264:
-audiocompression: mp3, 44.100khz, 96kbit stereo - you obviously want a small file
-videofilter: deinterlace drop field 2
-videofilter: resize to final resolution, bicubic
-videofilter: Donald Grafts smart-smoother, diam 5, str 25, inter no
-videocompression: divx5, 1-pass-quality-based, 70%, Psycho-Enh Strong
-save to AVI, takes ~2x of playing time on a 1Ghz-system
Why those unusual resolutions of 688x516 and 352x264? I don´t want to throw away resolution. PAL gives me some borders and those values have been veryvery carefully calculated to give me a 4:3-aspectratio with the highest possibly pixelcount.
I still have one problem: If recording anime, I get lots of ghost-shadows. Thats not a telesyn-problem, because telesyn and 25fps-PAL exclude each other. And it is not a interlaced-problem because this problem persists even when I drop field 2 and therefore no longer are using interlace. When I look closer, it seems to be an effect encoded right into the source-material at fullfame-basis - does that make sense?
My personal howto, which proved usefull in some local forums:
While googling the web I didn´t came across howtos or guides about this and tried my own ideas. But if you have some comments or places to look for, please participate in this thread.
You need Virtualdub, huffyuv, divx5 and some mp3-Codec. You might want to look into the Nimo-Codec-Pack, it brings many usefull stuff in one packet. Get Donald Grafts smart smoother
My workflow looks like this:
record using Virtualdub and huffyuv lossless compression.
cropping from 704x576 down to 688x550 to
-remove distortion at the borders
-keep a PAL-aspectratio of 5:4
-keep AVI-limits, xres 16bit-boundry, y-red at 2bit-boundry
this produces around 400MB/min at perfect quality. I wont explain segmented AVI, the 2GB or 4GB-limit, thats beyond the scope of this faq, ask your favourite searchengine.
I prefer a NTSC-ratio of 4:3 because thats the aspectratio of computermonitors and as divx is still a computeronly-solution. I want something which works outofthebox on a computer and this means a resolution of 688x516, 512x384, 352x264 of 320x240.
Always set audio and video to full process.
1-pass-quality-based at 60% to 80% gives good quality, but no control about the final size.
2-pass is harder and slower to use, but gives fine control about the final size.
Using "Psychovisual Enhancement Strong" seems to utilize the bitrate more efficient, I would say 3-10%. Thats work in progress, you might get a bloody nose in the future, but I still go for it.
to make an AVI at 688x516 or 512x384:
-audiocompression: mp3, 44.100khz, 128kbit stereo - videotapes can have an excellent audioquality
-videofilter: deinterlace mode blend
-videofilter: resize to final resolution, bicubic
-videofilter: Donald Grafts smart-smoother, diam 7, str 20, inter no
-videocompression: divx5, 1-pass-quality-based, 70%, Psycho-Enh Strong
-save to AVI, takes ~8x of playing time on a 1Ghz-system
to make an AVI at 352x264 or 320x264:
-audiocompression: mp3, 44.100khz, 96kbit stereo - you obviously want a small file
-videofilter: deinterlace drop field 2
-videofilter: resize to final resolution, bicubic
-videofilter: Donald Grafts smart-smoother, diam 5, str 25, inter no
-videocompression: divx5, 1-pass-quality-based, 70%, Psycho-Enh Strong
-save to AVI, takes ~2x of playing time on a 1Ghz-system
Why those unusual resolutions of 688x516 and 352x264? I don´t want to throw away resolution. PAL gives me some borders and those values have been veryvery carefully calculated to give me a 4:3-aspectratio with the highest possibly pixelcount.
Comment