Aspect Ratios (TMPGenc)

[migf1]

Too many things to look for, too little time

I didn't do my homework regarding DAR and SAR on the page I posted above yet, but after reboot's suggestions I did d/l Mainconcept MPEG Encoder and gave it a try.

I must say I was pretty amazed with the speed and quality! I did some tests with the default VCD/SVCD templates and the outcome was pretty darn good and fast!

I'm now d/ling the updated .pdf for the Advanced Settings, so I can start messing around and see what I can really achieve. Ultimately I'll try to figure out the analogy of settings between TMPGEnc and MCMEnc, so I can make some good comparisons (btw, any help that could save me some precious time would be greatly apreciated).

So, I think I'll push back a little bit the... study of that SAR/DAR page.

In the downside, MCPEnc is quite pricey for personal use and doesn't seem to have any extra filter capabilities. I plan to fully testing it until its demo period ends and then decide if is worth spending $150 on it. But the first impression was really-really... impressive!

Best regards,

Edit:
PS. If my memory serves me well, I think Adobe Premiere uses Mainconcepts MPEG2 encoders, doesn't it?

[migf1]

Ok, I'm back after a bit more extensive VCD testing, using both TMPGenc and MCMEnc.

I tried to push both encoders towards their quality limits, and I must say that TMPGEnc does a much better job (I remind you, I only tested VCDs).

My TMPGEnc settings, which can be found here: users.otenet.gr/~hkar/VCDSettings_PAL.zip (a zipped TMPGEnc template) take a bit more than a minute to encode each minute of 1/2 D1 PAL interlaced AVI sources (~22 minutes of encoding for a source of 21 mins).

In MCMEnc's "Advanced Video Settings" I used "Noise sensitivity: 1" and "Motion search mode: 11" (those settings yield to a "Search method: 11" and a "Search range: 8", in the Details window). The rest were left unchanged (they were already optimized by default, in analogy to my TMPGenc settings above... btw, I didn't check if both encoders use the same quant-matrices. I used their defaults).
It took MC about 17mins to encode the same 21 mins sources.

A quite noticable speed increase, but also a quite noticable quality difference. TMPGEncs output was always not only sharper but also smoother when the actors were moving.

But I may have missed some setting in MCEnc, since I'm quite new to it.

Best regards...

[duhmez]

I must recommend discarding every second field as a method of making vcd from d1 res, instead of merging fields. (This is for interlaced content, not film)
If you discard every econd field you get a tre progressive source for vcd encoding.

Put
separatefields()
selectodd()

to do this in avisynth.

[migf1]

I must recommend discarding every second field as a method of making vcd from d1 res, instead of merging fields. (This is for interlaced content, not film)
If you discard every econd field you get a tre progressive source for vcd encoding.

Put
separatefields()
selectodd()

to do this in avisynth.

Thanks once more for your hint dumhez!

If I got it right (I've never used AVIsynth) are you suggesting to deinterlace the source before the encoding? If yes, can I use TMPGEnc's built-in filter to deinterlace the source on the fly, during encoding? And if so, which method (Odd or Even)? I'm on a PAL system.

Best regards...

PS. Do you think it would be better if I capture from the beginning progressive 1/4 D1 (352x288) AVIs?

[duhmez]

What I posted was a different scheme for deinterlacing that produces fantastic results, but only works best on truly interlaced content. If it was originally a 24 fps film, don't do it this way. For many TV shows it's great however. Using this method, you would not set any deinterlace option in tmpgenc at all; you won't need to. You see, every odd field, is a 288 frame of a snapshot in time. 1/50th of a second later, the even fields are the 2/50th of a second. So all the odd fields together, is a 288 progressive stream, and every even set of lines is the same but a different time.

I suggest this: Capture 352,576 . then paste the script i posted intoan avs file (regular txt file with .avs extension instead of .txt)

This will make a movie with 50 frames per second of 352,288 , each fram in this case represents ONE of the fields. the selectodd() could just as easily be selecteven() both will work fine. by selecting odd or even, it will only feed the odd or even frame, and halve the framerate to 25 fps in the process, giving you a 352,288 progressive stream.

I sugest this becuse if you capture 352,288, whoo knows what method your capture card is using for the resizing; it may just be blending the 2 fields together, which for VCD is not desirable as the fields represent 2 different points in time. If, somehow, you KNEW your capture card was discarding a field, then capping an interlaced source right to 352,288 would be fine. Maybe see if you can get tech suport from your capture card company to get you these detaills, the engineers who make those chips know what they do and how they work, antuarlly.

once again, this is only true for content tht was recorded interlaced. You'd be surprised, some shows are progressive film source, you just cant tell because when you capture it you see the combing lines (especially us in ntsc world, 2 out of 5 frames is combed to hell when film is aired on tv, you pal people have it better)

To tell if it's intrlaced truly or not, easy peezy, drop the avi capture into virtualdub and analyze the video, if every frame ha the little combing lines in it, then it's interlaced.

[migf1]

Dumhez, you are a precious source of information. Many many thanks!

I will certainly try all your suggestions!!

Best regards,

PS1. I capture AVIs via a ViVo Pixelview (Prolink, in some countries) GeForce4 Ti4200/128Mb/AGP 4x graphics card.
Just in case anyone knows if it drops even (odd) fields when set to progressive cap or not.

PS2.I just bought a LiteOn LDW-811S dual burner. DVDShrink is the next proggie I have to test

PS3 I tested Satstorm's suggested (in another topic) software MPEG2 capturing capabilities of the MainConcept MPEG2 Encoder, and I was thrilled by the results. I capped 10 mins of standard TV broadcast, directly into 352x576 (vbrmin=3825, vbravg=6000, vbrmax=8000) DVD compliant video without dropping a single frame (P4 2.4B/533, 512Mb DDRAM/333). Total size=a little less than 500Mb. Burned it via Ulead Movie Factory 2.01 LE on the LiteOn bundle-included +RW. It looked good on my Philips Q40 standalone. Quick & easy!!!

PS4.This hobby is too... addictive

[FOO]

This is a hobby ? Oh shit I thought it was my job