meguIV: The Official Akiba-Online DVD Encoder (v1.0.1.1)

[Vitreous]

Thanks for taking it for a test drive. Glad to hear it worked well for you. k:

second pass: [IV] High Quality

Don't use this setting (it's the original MeguIV setting). It disables one of the best x264 features (B-Pyramid) giving less quality than the [Vit] settings.

i think i can see a slight drop in quality and lines seem to blur a little under quick movement

There isn't any drop in quality. Almost exactly the same processing is done, it's just done faster. To be absolutely precise (and to be clear that this kind of thing is quantifiable) the only differences are:
- QuickTGMC doesn't internally pad the image by default, which can create tiny differences (1 pixel or so) on the very top and bottom lines, but typically fixes a couple of problems there in doing so.
- There are a couple of blurs used internally in the algorithm that are done with ~0.4% less accuracy
- There is a possibility of an ~0.4% overshoot in the resharpening
- There is a bug-fix in the motion analysis, which makes actually things better
I test on a range of sources, comparing before and after: either synchronized side-by side for motion, or flicking between 400% zoomed fullscreen versions for static. There is simply no appreciable difference between your settings and MeguIV standard.
You really get something for nothing here. :grassdance:

EDIT:

what are the three [Vit] audio presets equivalent to in kbps?

Those settings use AAC variable bitrate (VBR) - it's a close equivalent to the approach used for the video: set a quality level rather than a bitrate, then the encoder uses whatever bits it needs to reach that quality level. This graph shows the estimated bitrates for a given Q-value (0.3 ~ 96kbps, 0.4 ~ 128kbps, 0.5 ~ 160 kbps):
[fullattach]333705[/fullattach]

 

[isityours]

...It disables one of the best x264 features (B-Pyramid) giving less quality than the [Vit] settings.

maybe that was why i thought i saw a slight difference in picture quality. i will use the default preset from this day on. thanks.

also, i added a question about the audio presets but i edited the original post so you may not have seen it.

 

[Vitreous]

maybe that was why i thought i saw a slight difference in picture quality.

No, you actually used the same setting that was in original MeguIV. By chance you used the most equivalent possible settings. (edit: unless I've broken something recently... hurries off to test worriedly)

Edit2: I should note that the x264 encoder tends to give a different "grain" across the whole image even if there is just the tiniest pixel or two of difference. It's not noticable in motion, but it is when comparing static frames. So comparing x264 output (and hence the final MKVs) can be rather subjective. You can compare TGMC against QTGMC by looking at those huge uncompressed intermediate avi files (which are deleted unless you change the settings in megui).

See my also edited post for audio info.

 

[no__One]

Only to tell you, why I remain so silent and couldn't make tests.

Currently, I am on a netbook Atom... and trying to fix my main PC... :study:

Hope to be back to work and give a try to Vitreous'S + MeguIV improvements by the end of next week. :scared:

 

[Vitreous]

I've updated to meguIVit version 0.1.3

Ensure you delete the Sandbox before installing this one.

The update contains a range of preset changes:
- The AAC audio presets have slightly tweaked Q values and I've put the estimated bitrate in the preset name. I've also added a higher quality preset (~190kbps)
- I've added [Vit] MP3 presets. They use VBR encoding - the same quality-driven approach as the AAC presets, discussed at the end of this post. Again the estimated bitrate is given
- AAC should be preferred unless you have a compatibility reason otherwise

- I've tweaked Rollyco's original MeguIV "[IV] High Quality" x264 preset. It's OK to use now as I re-enabled B-Pyramid. But it only uses 3 reference frames so it provides a tiny bit more speed than the [Vit] default, but loses some quality in places.

- After some testing, I now recommend against any speeding up of the x264 preset. Maybe it's because we're still using an older x264 encoder (for now), but the quality loss when speeding up the 2nd pass is worse than I remembered.

 

[Konata Izumi]

MeguIVit 0.1.3
ERROR during Pre-Rendering "Super Fast" One-click Setting
____________________________________

I'm getting 13.37FPS using [Vit]30fps E: Medium [One-click Preset] during Pre-Rendering job.
I'm getting 15.75FPS using [Vit]30fps H: Very Fast [One-click Preset] during Pre-Rendering job.

AMD PhenomII X2 550 Black Edition 3.10Ghz
896MB RAM (DDR2)

I love you Vitreous :grassdance:

 

[Vitreous]

Returnil Virtual System 2010 flags a Megui.exe inside the sandbox as MALWARE.

I had a virus alarm last year from meguIV - after submitting for analysis it came back as false positive. I've also just today had an alert on Megui.exe in the STUBEXE folder, so I've submited it for analysis again. You should do the same. I believe it's to do with the Xenocode packager that Rollyco uses, which runs meguIV in a sandbox. This allows meguIV to run in a portable manner without fully installing AviSynth ior Megui etc. Unfortunately, this also seems to create suspicious looking files

I'll have a look to see if there's a better way to package.

ERROR during Pre-Rendering "Super Fast" One-click Setting

Hmmm. I thought I'd fixed the Super Fast / Ultra Fast crashes. Obviously not. Only seems to happen to me in the Xenocode sandbox, although that might be coincidence. Again, I'll have a look.

15.75fps is still pretty good for that machine though!

 

[xater]

I'm using Vit's script but with vanilla megui...:scared:

Mainly for old rips, but recently I tried in a couple of full DVDs, Saaya's Love16 set1, which I made it to 60fps (never had tried that before), used QuickTGMC( Preset="slow", Sharpness=0.35) and the result is pretty good although the iso looked good too... on the other hand, I also did Michika Fujii's IV, which was widescreen black bar-ed to 4:3 and looked horribly combed while playing it... used "faster" on it and only 30fps and the result was also much better... k:

But I guess I'll go back to re-rips -no interesting isos atm for me-, what does the InputType= numbers do? I did a wmv re-rip, and when trying with =1, it got ugly, worked with =2, but I would like to know what to use (I guess =0 is for interlaced stuff, but the others?uzzled

 

[Vitreous]

I'm using Vit's script but with vanilla megui...

Glad to hear it. Best way for those of us who like to experiment.

Sharpness=0.35

This reminded me of something. This sharpness is the original meguIV setting: it's fairly soft given that the default sharpness is 1.0. I should mention for the meguIVit users that its presets use the default 1.0 sharpness setting, which I forgot when mentioning differences in an earlier post.

what does the InputType= numbers do?

InputType=0 is for standard interlaced material
InputType=1 handles any kind of progressive material
InputType=2,3 are difficult to describe. They assume that you have progressive material that has been deinterlaced, but that the deinterlacer did such a bad job creating the new odd / even lines that they're not worth keeping. Instead it will rebuild the vid from the other lines which are hopefully still close to the original lines from the source material. So these modes discard the poor lines, then reweave the original lines back into a interlaced stream again (now 15fps). It then runs the normal QTGMC process on this stream to give a 30fps result again... (!)

The practical difference for a rerip is this: mode 1 will give a bit more detail, but will occasionally have slightly jagged edges where the original deinterlacing was poor. Modes 2,3 give a smoother result but will lose a little detail due to the discarded lines. One of mode 2 or 3 will be better than the other, usually 2, it depends on whether the odd or even lines are closer to the original source. It's can be difficult to work out which of mode 1, 2 or 3 will be best, so you may need to make a test rip of each. However, if you see lots of residual combing in the source, you might be better with 2 or 3. I've most often gone with mode 2.

 

[Vitreous]

Returnil Virtual System 2010 flags a Megui.exe inside the sandbox as MALWARE.

Avira have confirmed this as a false positive.

I had a quick look and with some tricky file organization, MeguIV could just be packaged in a zip to avoid these alerts - only requirements would be .NET (ubiquitous) and the ffdshow HuffyUV codec. Not inclined to actually create this atm though

ERROR during Pre-Rendering "Super Fast" One-click Setting

I can't reproduce any errors with this setting. Could you try again, and when you get the error go to the "Log" tab of MeguIV, right-click and choose Save>>Log. Then post that log here so I can investigate. Thanks

 

[astrayred]

I'm using Vit's script but with vanilla megui...:scared:

Mainly for old rips, but recently I tried in a couple of full DVDs, Saaya's Love16 set1, which I made it to 60fps (never had tried that before), used QuickTGMC( Preset="slow", Sharpness=0.35) and the result is pretty good although the iso looked good too... on the other hand, I also did Michika Fujii's IV, which was widescreen black bar-ed to 4:3 and looked horribly combed while playing it... used "faster" on it and only 30fps and the result was also much better... k:

But I guess I'll go back to re-rips -no interesting isos atm for me-, what does the InputType= numbers do? I did a wmv re-rip, and when trying with =1, it got ugly, worked with =2, but I would like to know what to use (I guess =0 is for interlaced stuff, but the others?uzzled

Regarding Michika Fuji's IV, I should probably let you know that I did a 60fps TGMC'ed, cropped black bars, and upscale to 720x480 16:9 using NNEDI3 rip. I can assure you the quality is nothing short of astonishing, it's virtually indistinguisable from a original 16:9 video. Seeing as I used extreme settings on it (NNEDI3 TGMC, x264 @ Placebo), I should probably get around to sharing my computer's hard work with everyone...

Irie Sayaa's video was a bitch. It was already progressive to start with and letterboxed to boot. At least with Michika's video being interlaced I still can fix it, but how do you expect me to fix this POS?

Sometimes I swear the DVD authoring people are idiots. Oh well, I'm not the first one to have said that

 

[xater]

uzzled:

Are you talking about this one?



I don't recall seeing black bars in that one... :dunno:

 

[Vitreous]

Seeing as I used extreme settings on it (NNEDI3 TGMC, x264 @ Placebo)

I haven't been convinced by NNEDI3 yet. It seems to make as many new mistakes as fix old ones. Although I haven't experimented with the parameters too much, have you?

It was already progressive to start with and letterboxed to boot.

QTGMC InputType ftw...

I should probably get around to sharing my computer's hard work with everyone...

Yeah, payback time! ayup:

 

[Konata Izumi]

I've successfully finished a 2hr video (MXGS_264.iso) encode using Very Fast Preset just now.
I got a 1.55 GB encode and am not happy with the filesize



Is it alright to specify filesize at the One-Click settings?
How can I make a proper 23/25 fps encode?

Could you make a script/preset for a 23.976/25.976 fps encode?

 

[Vitreous]

Is it alright to specify filesize at the One-Click settings?
How can I make a proper 23/25 fps encode?

[Edit2 - I keep changing my mind about this answer]
As this site concerns Japanese content we're focused on 30fps NTSC and there are no presets for 24/25 fps. But if I made presets for 24/25fps with the same quality as the 30fps settings then your file would get larger. Because when you show less frames per second then each frame needs to be higher quality to give the same overall quality impression.

However, you can easily adjust the presets to make the file smaller if you wish:

- Choose the One-Click setting as normal, for example "30fps F: Fast"
- Adjust the Video Profile, choose "480p 30fps A Bit Smaller" to reduce the filesize a little
- To reduce filesize even more use the "480p 30fps Much Smaller" setting
- I would guess the "A Bit Smaller" setting would get you 1.3Gb and the "Much Smaller" setting would give you 1.1Gb.
- meguIV is designed for quality rips, so if you need even smaller then you will have to go into the "Video Profile" config and choose your own settings.

Also, note that the faster presets usually make bigger files.

 

[Konata Izumi]

• Can I use [HQ] tag when posting here if I use MeguIV with [Vit]30fps H:Very Fast - 480p 30fps Much smaller presets?
• The "much smaller" presets would make the encode faster or slower?
  
• When will you update the x264 encoder?
  
• my system became sluggish during 2nd pass compared to the pre-rendering/1st pass is this normal?

Can anyone give me seeded torrent/DDL links to ISOs that doesnt have any rips available for download yet? maybe I could rip it. :tea:

 

[Vitreous]

# Can I use [HQ] tag when posting here if I use MeguIV with [Vit]30fps H:Very Fast - 480p 30fps Much smaller presets?
# The "much smaller" presets would make the encode faster or slower?
# When will you update the x264 encoder?
# my system became sluggish during 2nd pass compared to the pre-rendering/1st pass is this normal?

1) No, 2) Not sure - just try it, 3) Don't know, 4) Read step C

Please don't ask questions (1 & 4) which are already answered in the main meguIVit post.

 

[astrayred]

I haven't been convinced by NNEDI3 yet. It seems to make as many new mistakes as fix old ones. Although I haven't experimented with the parameters too much, have you?


QTGMC InputType ftw...


Yeah, payback time! ayup:

NNEDI3: I think it works better for upscaling than deinterlacing. I would agree that the speed slowdown is not worth the quality increase (if there is one) that one would get using NNEDI3. I would say that however, at this point I'm pretty much a script kiddie. Everyone on D9 raves about the quality and I buy into the hype

@Xater: You know it's weird, I went looking for more Irie Sayaa videos and came across an ISO of the cover you posted. It's exactly the same as the one I'm talking about, except that mine is letterboxed! I got it from a torrent I think...

Vitreous: Sometimes when upon playing a video in DGIndex, it says frame type interlaced but a visual inspection of the video itself reveals no interlacing lines. Is it possible that there is a flag that one can set to interlaced and progressive, and in this case it's a progressive video flagged as interlaced to create some sort of "fake-interlace" for whatever reason?

 

[Vitreous]

Vitreous: Sometimes when upon playing a video in DGIndex, it says frame type interlaced but a visual inspection of the video itself reveals no interlacing lines. Is it possible that there is a flag that one can set to interlaced and progressive, and in this case it's a progressive video flagged as interlaced to create some sort of "fake-interlace" for whatever reason?

I don't know. I've only observed the same as you: that some DVDs have progressive content that says that it's interlaced. In every case I've seen I'm sure that the content was originally interlaced, but for some reason has been deinterlaced prior to authoring. In some cases the deinterlacing that's been used is terrible. I'm guessing it's a mistake made during the editing process, as often the progressive/interlacing changes scene-by-scene. Maybe one of our resident experts has some insight?

 

[Rollyco]

http://neuron2.net/dgmpgdec/DGIndexManual.html#Video

Frame Type - Displays Interlaced or Progressive depending on the value of the MPEG2 progressive_frame flag. Please be aware that this merely describes how the frame was encoded; it says nothing about whether the content of the frame is progressive or interlaced. It is common for progressive video to be encoded as interlaced, and vice versa.

The ONLY reliable way to tell if a DVD video stream is interlaced is to apply SeparateFields() and framestep. Manually.

In-depth explanation, see Annex K:
http://neuron2.net/library/mpeg2/eip.pdf

Mile-high overview of MPEG2 headers:
http://dvd.sourceforge.net/dvdinfo/mpeghdrs.html