im@s2 manual blue screening - A quick adventure and some notes.

[Setsuna]

Well, I decided (Well, it is a project of mine) to give the bluescreening a shot. I figured that someone else may ask, and it might prove to be a useful set of notes for those wanting to follow along.

The end result is here:

http://youtu.be/3kE1jzsH3Mw

For those who don't know - it's a method of using algorithms (Edge + difference) after doing composite comparisons on several backgrounds in an attempt to remove the background. It's not perfect, but it leaves a lot less stuff to manually mask out.

For those who don't know me, I have about as much Japanese knowledge as I'd expect most people here to understand forensic accounting. Namely, next to none, so the only way I've managed is via trial and error.

http://www.nicovideo.jp/watch/sm14364377

Even though Javie is mostly in Japanese, the NukIM@S-3 addon is... mostly in english, oddly enough.

You can get it here.

http://sourceforge.jp/projects/javie/

Requirements for this are:

- Windows XP SP3  or better.
- A Nvidia or ATI card with either Catalyst 11.2 drivers, or GeForce 260.99 or better (Depending on what your graphics card is of course)
- Java version 5 or better http://www.java.com

Things you might want on hand:
- Lots of time. It took me most of about 6 hours to get the demo done. Granted, I'm not a video editor and I imagine real ones will post less time.
- A good video editor program (Which you will need for one of the steps, and of course, processing afterwards)


Of course, being me, I try something that's clearly not on the 'For beginners list' for my first attempt.

The song is fairly simple. i...

Okay, bad jokes aside, the difficult part was in the setup, because I picked someone who sort of melds in blue screens (Azusa) and I picked a... mostly blue costume set, and I shot it exclusively at long range, making it very difficult to apply anything other than the basics.

That and I tried to make a uniformed blue screen setting over a 2:15 song, when in all fairness, I should have chopped it up and done a better job every time I should have changed the settings.

However it's a demo, not a production reel, and it's mostly a proof of concept.

In the process you will learn a lot of things, which you will care about when you start this.

1. You require at least two (up to 4) videos to attempt this masking method. You also want to keep around a couple of spares.

2. You'll need to make sure that each of the reels are IDENTICAL in their dance sequence. This sounds easy. It's not. Here's why:

- You'll discover that there are actually FIVE breeze settings. (two in each direction, and a neutral.) This isn't actually a bad thing, since it adds a little bit of variety. Thing is, if you're dealing with Miki's hair, you'll lose about half of it because her hair won't bloody well be in the right spots anymore.

This means you need to be VERY particular about the stages used. I'll be adding to the compendium general notes a breeze measurement now, because of this.

- You'll also discover that you'll need the lighting to be identical all the way through. It crosses off most of the indoor stages, which would have been useful. Once again, time to be particular folks.

- Did I mention the reels need to be the exact length and start at the exact same point and precision is basically queen here? If I didn't, I mean it. this is why you want a good video editting program, to prep your video. As well, don't try this with youtube rips, you WANT the true 59.94 frames that im@s outputs. (Then again if you're using youtube video to do serious video work, you should be clobbered on general principle.)

- You'll have better luck doing short spurts (not whole clips), and/or working closer up in the camera angle. Remember that you'll need to prevent the bounciness, and you probably (for sanity's sake) keep it to a single angle per run. You get better results if you're zoomed in (at Mid or close) but remember to keep the angles and any bounciness consistent.

Other than that, have fun.

Use the tutorial (slow it down and basically copy it. Yours will be in english, and they're in the same order as the Japanese version) and remember that the trick is the bits in red when you switch to the Difference + Edge (Filtered) Mode are the bits that will 'show up' in the blue screen.

I won't walk through all the steps (because the Japanese tutorial can probably demonstrate what you're gunning for better than I could describe in words) although remember to change your output to 'Result from Difference + Edge (Filtered) to  'stage X with unsharp mask' with X being the number you want to stop at.

And then of course, output the AVI with encoding settings your computer can hack. Just bear in mind that depending on your computer, you could be there a while with this...

[animagic4u]

Thank you.
Although I have a lot of knowledge in the way of video editing this seems to be a different area completely.

Though, aren't they going to release a m@d patch for ps3 (lyrics remove and blue screen)

[TheTanStar]

Subtitles can just be removed by pressing R3. This process has its own limitations; most of which Setsuna covered. However, I should add one more: Particle effects. If you look closely at the demo video at the end, blue particles are visible across the characters. Too bad there isn't a way to disable them that I know of.

[Setsuna]

Subtitles can just be removed by pressing R3. This process has its own limitations; most of which Setsuna covered. However, I should add one more: Particle effects. If you look closely at the demo video at the end, blue particles are visible across the characters. Too bad there isn't a way to disable them that I know of.

I forgot about that one. I was pretty tired when I wrote it up.

You can apparently dodge some of those, but it requires very specific reels, and a lot of time to manually add it back in. Basically, have them with different effects, and since the two will displace at different points.

I didn't put that much time, since for some odd reason all the stages I had kept spitting confetti all in the exact same spots.

Thank you.
Although I have a lot of knowledge in the way of video editing this seems to be a different area completely.

Though, aren't they going to release a m@d patch for ps3 (lyrics remove and blue screen)

No word on the blue screening process yet (Although subtitles are gone with a click of a button). I thought considering that (and my comments about how im@s 2 is slowly replacing im@s L4U as the tool of choice) that I cover whatever ways Cael.K (I didn't get the chance to find it myself) dug up concerning manual bluescreening methods and quickly demo them for the fun of it.

[Cael K.]

Glad it helped!

[Nii-san]

I'm interested in making my own videos also, but I just wanted to make sure of a few things if anyone could reassure me on them:

I'll need a capture device, something like Blackmagic Intensity using the HDMI, correct?
Also from what I understand, I can use Adobe Premier Pro to capture in full 1080p, After Effects for special effects, and Photoshop for screenshots.

I just wanted to be 100% sure this was the way it was done. Thanks! 

[TheTanStar]

HDMI recording will not work on a retail PS3, only a dev unit (because HDCP can only be disabled on a dev unit), at least until a new CFW comes out that can disable HDCP. In other words, the best way to record is through component + SPDIF/TOSLINK for 5.1 audio. Also, DO NOT record in 1080P, it's usually an upscale, making the output just look worse than it needs to be.

[Nii-san]

HDMI recording will not work on a retail PS3, only a dev unit (because HDCP can only be disabled on a dev unit), at least until a new CFW comes out that can disable HDCP. In other words, the best way to record is through component + SPDIF/TOSLINK for 5.1 audio. Also, DO NOT record in 1080P, it's usually an upscale, making the output just look worse than it needs to be.

Oh really? Man... but you're right, I forgot that iDOLM@STER 2 is 720p (at least that's what it says on the back of the case)

So I should record in 720p using component and 5.1 audio using toslink? Which means I'd need a capture device with component and toslink in/out...

I skimmed the internet and recording 5.1 using toslink doesn't look promising, hardware and software wise, so maybe just component 720p with stereo audio?

I just wanted to record it as lossless as possible...but 720p stereo is good enough huh? 

[Setsuna]

HDMI recording will not work on a retail PS3, only a dev unit (because HDCP can only be disabled on a dev unit), at least until a new CFW comes out that can disable HDCP. In other words, the best way to record is through component + SPDIF/TOSLINK for 5.1 audio. Also, DO NOT record in 1080P, it's usually an upscale, making the output just look worse than it needs to be.

Well, actually, im@s 2 X360 isn't, but the gains in doing so are not worth the trouble (And it requires overclocking it due to the fact there's a non zero chance it'll literally have framerate issues.)

The bigger problem is that recording in 5.1 audio requires a lot of work (mostly due to the recsyncing since you can't use the capture card to handle it.)

As well, Dev units disable HDCP only when a game's running - it's a technical point, but mostly a note you can't use it to output 1080p60 (or even 3D) out of a component for blu-ray. (If you've ever considered sticking it to the MPAA or the RIAA and streaming the Titanic in 3D, you'll have to work at it a bit harder at it, unfortunately.)

In any case, recording in lossless 1080p60 (or 3D while we're on the topic) isn't really feasable - the cost of the cards alone to attempt it are somewhere in the ballpark of a couple of thousand dollars, and you still need to add further support units, batting it well in the vicinity of about $4000 from my last glance at it.

Oh really? Man... but you're right, I forgot that iDOLM@STER 2 is 720p (at least that's what it says on the back of the case)

So I should record in 720p using component and 5.1 audio using toslink? Which means I'd need a capture device with component and toslink in/out...

I skimmed the internet and recording 5.1 using toslink doesn't look promising, hardware and software wise, so maybe just component 720p with stereo audio?

I just wanted to record it as lossless as possible...but 720p stereo is good enough huh? 

For most part, yes. Most of the time for MAD work, you won't be recording the audio and will be superimposing on the top. If you have a 5.1 track or better to superimpose, do that.

Finding something to take it isn't that hard an issue - putting it all back in sync on the other hand requires some fiddly work. Most people don't expect youtube to fire 5.1 regardless anyway.

If you want real lossless recording, you'll need to record every pixel unmodified. The Blackmagic Intensity PRO (you can't use the Intensity - no component, and unless you own a dev unit (or somehow can hack it in) you can't use the HDMI port due to HDCP) will most likely be sufficent.

However, you'll need to have a HD fast enough to keep up with the approxmently 130MB/s. Most people use several HDs in a RAID0 format which works most of the time (some issues eventually of course) . It's cheaper than my method, but I don't settle for near enough - I use a SAS HD to handle that.

[Nii-san]

Well, actually, im@s 2 X360 isn't, but the gains in doing so are not worth the trouble (And it requires overclocking it due to the fact there's a non zero chance it'll literally have framerate issues.)

The bigger problem is that recording in 5.1 audio requires a lot of work (mostly due to the recsyncing since you can't use the capture card to handle it.)

As well, Dev units disable HDCP only when a game's running - it's a technical point, but mostly a note you can't use it to output 1080p60 (or even 3D) out of a component for blu-ray. (If you've ever considered sticking it to the MPAA or the RIAA and streaming the Titanic in 3D, you'll have to work at it a bit harder at it, unfortunately.)

In any case, recording in lossless 1080p60 (or 3D while we're on the topic) isn't really feasable - the cost of the cards alone to attempt it are somewhere in the ballpark of a couple of thousand dollars, and you still need to add further support units, batting it well in the vicinity of about $4000 from my last glance at it.

For most part, yes. Most of the time for MAD work, you won't be recording the audio and will be superimposing on the top. If you have a 5.1 track or better to superimpose, do that.

Finding something to take it isn't that hard an issue - putting it all back in sync on the other hand requires some fiddly work. Most people don't expect youtube to fire 5.1 regardless anyway.

If you want real lossless recording, you'll need to record every pixel unmodified. The Blackmagic Intensity PRO (you can't use the Intensity - no component, and unless you own a dev unit (or somehow can hack it in) you can't use the HDMI port due to HDCP) will most likely be sufficent.

However, you'll need to have a HD fast enough to keep up with the approxmently 130MB/s. Most people use several HDs in a RAID0 format which works most of the time (some issues eventually of course) . It's cheaper than my method, but I don't settle for near enough - I use a SAS HD to handle that.

Alright so the Blackmagic Intensity PRO is a good choice...

Regarding the hard drive, what kind of specs would you recommend?

Not sure if my main hard drive would be fast enough, it's a Western Digital WD5001AALS-00L3B2. Looking around, it doesn't seem to reach reads of 130MB/s....

I assume a solid state drive would be more than enough? Again, looking around, the Corsair Force 3 looks good...

Thanks for the advice! I appreciate it

[Setsuna]

Alright so the Blackmagic Intensity PRO is a good choice...

Regarding the hard drive, what kind of specs would you recommend?

Not sure if my main hard drive would be fast enough, it's a Western Digital WD5001AALS-00L3B2. Looking around, it doesn't seem to reach reads of 130MB/s....

I assume a solid state drive would be more than enough? Again, looking around, the Corsair Force 3 looks good...

Thanks for the advice! I appreciate it

I'll note that the only reason you'd actually WANT to record every pixel unmodified is so you can encode it specifically later. This takes time. It's considered best practice to capture at the highest possible, then encode to suit, but it doesn't help if to do this it'll take you two days to do a 2 minute video.

It might be practical to record a compressed image, and start from there.

For an idea, I run a i7970, which is a six core monstrocity and reencodes a raw video to 30mbps stream from the 14.7GB raw to 500MB in the space of about 20 minutes per. The video in question is 2 minutes 15 seconds. You could do better now, but it's still outside the realm of 'cheap computing'.

(It takes me about an hour to upload to youtube, which is sort of ironic, but that's net connections in Australia for you.)

Your computer mileage may vary (and most likely will, since you'd need something made in the last year to get anything close to those encode speeds.) and it depends who, where and how you plan to distribute. Youtube isn't exactly demanding of its high res specifications, for example. My freelance work on the other hand usually demands a minimum standard far exceeding youtube's distribution (about an order of magnitude or two.)

Should have mentioned that, but I forgot. Long day today.

Anyway, on the subject of hard drives and other bits and pieces.

Bear in mind that when it comes to write speeds for capture devices, you NEVER talk about burst. Sustailed write is the only write that matters, simply because they don't ever stop until you tell it to stop recording.

SSDs can provide the speeds for raw writes but read the below first.

I'll actually outline a couple of points over all the solution types available.

Scalablity:

SSDs are not very scalable - Getting one or two may be okay, but for an idea what a SAS will get you, you can get a 600GB SAS for about 600 AU, and that SAS will clock in at 190MB/s until 99%.

Granted, you'll need a SAS controller, and that'll range from maybe 400 to the moon, so bear that in mind with the cost calculations. The SAS controller's main function is to cache a lot of data and make sure if the drive ever hiccups, that the write commits still keep pace.

Getting five SSDs to get roughly the same figure (assuming 128GBs per RAIDed up) is significantly more expensive. If you think you can limit your records to 4-5 minutes, a 64GB SSD would be okay. You'd need 2 for 10 minutes, and so on, but eventually the linear catches up with the lower cost per GB of a mechanical drive.

Enough standard HDDs in a RAID0 configuration is cheaper than BOTH (I mean 400 dollars will get you what, 2-4 1TBs with good write speeds and NCQ disabled) and even if you add in a dedicated RAID card (which is recommended and can be gotten for sub 50 dollars) you'll find that you'll only maintain 150-140MB/s until the entire array is half full.

Then again, if you built a 4TB array, uh, 2000GB is plenty for recording at 720p. It does mean you'll need to enforce cleaning up disipline every so often though, but if you ever wanted to do a lossless LP record of im@s2, you'll probably have the space to get quite a few hours out of it. There's also more risk of a write 'hiccup' and you'll lose data during that fraction of time.

However, you're completely bottlenecked at 4 drives (since RAID0 doesn't get you any faster beyond the 4.) so if for some reason you want to record at 1080p30 or better, you'll have to look into something else. (Namely SSDs or SAS)


Lifecycle:

A SSD lifecycle is measured in number of commits (somewhere between 300000 and 500000 commits usually.) as the NAND becomes less receptive and more worn as you write to the drive. If you're planning to record lots, bear in mind a SSD constantly writes to the drive during the record process. it's more than 1 commit per record. (I can't give exact figures though.)

If you're doing it as a hobby and use it as a scratch drive, you might replace it when the drive naturally falls over in 2-3 years, and if you can keep the record times within spec, it's a relatively cheap reliable option. I have known someone who managed to wear the NAND down in 9 months though, but it was a production recorder in near constant use, so that's the extreme case.

A SAS drive is rated somewhere usually between 1 and 1.5 million hours of operations. These things don't come with 5 year warranties for nothing - They're expected to actualy last that long, and businesses demand it.

HDDs are somewhere inbetween. They're also readily available to replace when they break down.

-----

Anyway, that'll give you the rundown on some of the specifics. There's also very technical issues, namely if you REALLY want the two extra bits of colour, which will determine if the BMI Pro is good enough for you or not... But I have no idea what you have in mind.

[Nii-san]

Ah sorry for not saying what I'm actually planning on doing hahah 

I was just looking to save my performances and share them (on YouTube...maybe NicoNico also?..so I can link back to them on various sites) in HD (I'll be happy with 720p and nice stereo). Nothing professional, at least not for now, since I hardly know anything about all of this. I just wanted to start out with simple capture/encode/share, without doing anything like blue screening, like what you've been doing. Once I get my feet wet with this whole process, maybe I'll look into more advanced video editing; the only video editing I'll look into doing at first is probably just adding like a simple watermark.

So I'm not looking to get perfect lossless quality if it's going to be that expensive, time consuming, and difficult 

I was looking at it as more of a hobby, spend a couple hundred dollars, with really good quality, cause I'm really picky with audio and video quality...I want my idols to look and sound good! 

[Setsuna]

Ah sorry for not saying what I'm actually planning on doing hahah 

I was just looking to save my performances and share them (on YouTube...maybe NicoNico also?..so I can link back to them on various sites) in HD (I'll be happy with 720p and nice stereo). Nothing professional, at least not for now, since I hardly know anything about all of this. I just wanted to start out with simple capture/encode/share, without doing anything like blue screening, like what you've been doing. Once I get my feet wet with this whole process, maybe I'll look into more advanced video editing; the only video editing I'll look into doing at first is probably just adding like a simple watermark.

So I'm not looking to get perfect lossless quality if it's going to be that expensive, time consuming, and difficult 

I was looking at it as more of a hobby, spend a couple hundred dollars, with really good quality, cause I'm really picky with audio and video quality...I want my idols to look and sound good! 

Nico requires you to pay to upload video. Considering you need to make various arrangements, it can prove tricky.

My role is sort of unique - I don't actually do video editting, since most of my work is based around proof of concept as well as support and supply services. I'm not a MAD maker, but I support a few of them.

Well, to be honest, you might (key word is might) manage with a compressed type. They can vary from the Hauupage PVR (Although I don't personally like it) and a few others which is about 300 US, to the Intensity H264 encoder (but that's 500 dollars but it takes everything but the kitchen sink and can be carried around in your pocket, but it's a quality piece of on the go equipment.)

It really depends on your hardware and what slots you have available.

I don't use a mac, but if you have one with thunderbolt (once again, a recent thing) I've been advised by a few people I know they'd marry their thunderbolt enabled Intensity Extreme, only because it's more reliable than the USB3 Intensity Shuttle for lossless work, but still has the 'pick up and go' qualities and can be used with their SSDs for lossless on the fly. That's about 300ish as well, from my knowledge.

Generally at the entry level of pixel perfect recording, you're almost always looking at a PCI-e card (Short of using the Intensity Extreme)

The most efficent is the Blackmagic Intensity Pro (And at 300 dollars it's cheap!), and for basic 3D and 1080p/2k you're looking at the Extreme 3D (1k). At the top end of town (at several thousand dollars) AJA are pretty much king of the mountain, and they also (for the low low price of just 5000 US) sell a converter which literally changes one signal to any other single you damn well please while giving HDCP the finger.

There's others (including I think Avmedia who do a suprisingly good RGB/D-sub solution which I've worked with and is very nice, if PC work is your deal, but it's expensive. Research your options, particularly if you want to record classic (240p or lower) or Arcade games.) although I stopped playing at the low end of town years back. The BMI was when I got it back in 2007 a godsend (although at 600 US, which back then converted to about 800 AU. Ouch!)

... but don't mind me. I really would like that swiss army knife of signal converter hardware, and I enjoy playing with video, particularly in the creative and the big leagues. Being able to stream in HD has a few other advantages too, particularly when it comes to having people help you play Japanese RPGs in, well Japanese.

This is all assuming you have a computer that's in the last couple of years. You might struggle with anything older than a quad core (I remember when a good 3000kbps encoded video took 8 HOURS per run, and had to do it twice for the two pass and the video length was somewhere in the vicinity of 5 minutes) simply because of the time required to encode all the raw data.

Now, if you have the patience you can live without a faster computer, or if you have a spare which means you can leave your encodes to run overnight while you do your work.

If you don't have a PC that won't drive you nuts (or you don't have one with a PCI-e slot on hand), and you still want in, personally I would advise (if you could afford it) a Sandy bridge i7 computer, well configured (which will set you back about 1000 all up and built) and load a BMI into that, and maybe a SSD or a RAID card + several HDs (not sure which one's cheaper).

If you don't like it enough, at least you have a computer that'll do well for a while yet, and you can play FPSes, Starcraft 2, Diablo 3 or something.


As for programs, I can't condone piracy in any manner, but to start, try using virtuadub. It's fairly good as a starting point, and then you'll need to find a good h.264 encoder. Unfortunately the options I have (Carbon Coder for batch encoding and watch folders and Adobe Premiere Pro) are uh, in the vicinity of several thousand dollars and are probably not a good idea unless you can acquisition them.

[Elixir]

Nico requires you to pay to upload video.

No.. it doesn't. Premium is only for people who want to browse in peak times without falling into economy mode, watch streams set to premium users only, and events that only premium users can view. I've uploaded videos to nicovideo before.

http://www.upload.nicovideo.jp/upload

I'm not sure why you would want to, though... Youtube has already surpassed Nicovideo in many ways.

[Setsuna]

No.. it doesn't. Premium is only for people who want to browse in peak times without falling into economy mode, watch streams set to premium users only, and events that only premium users can view. I've uploaded videos to nicovideo before.

http://www.upload.nicovideo.jp/upload

I'm not sure why you would want to, though... Youtube has already surpassed Nicovideo in many ways.

Ah, I thought it was a premium thing. The more you learn I guess.

But yeah, apart from its odd status in Japan (as some sort of informal hub things that would normally get people into IP trouble. Japan's a bit strange when it comes to IP enforcement.) there isn't much from a technical perspective.

Then again, since nico isn't in the US, it might pay to keep it around, in case youtube (which is subject to US law) suffers another set of screws curtesy of the US government.