[kadajawi]

have you ever taken a telecommunications class? have you ever been paid to shoot video? how about posting up some samples of video that you have shot and edited... lets see some of your hand-held dslr work.

there are a lot of hot-air opinions out here, with only one or two people posting up anything that they have shot.

where is the proof that any of you can pull off a hand-held dslr shot? for example, that grispie shot of the cat was pretty good!

"hand-held" is when you hold the camera *only*, nothing else, and no, that's not how pros shoot video... you'll see, when you actually try do it, instead of just talking about it.

I have uploaded 2 videos I have shot on the general K-3 video thread in this forum. The link is in the post right above yours. And no, those aren't edited, they are raw footage meant to demonstrate the SR system of the K-5, and how well it (IMHO) works. They are also meant to document my last China trip, but that's more for private purposes. I'm quite happy with what I shoot handheld. It's not perfect, but for my purposes more than good enough. I wouldn't be happy with K-01 footage, or, probably, K-3 footage. K-5 though, yeah I am.


Would I do paid work completely handheld? Probably not, while I could probably get good enough shots (depending on the purpose), I may not be able to get them every single time.


What YOU said was use a tripod, that's what pros use. And, I'm sorry, but THAT I can't agree with. Sometimes, yes, tripod. Sometimes a dolly. Sometimes a steadicam rig. Sometimes it's shoulder mount. Or perhaps even handheld. If you had said they shoot stabilized in some form I would have agreed. But only tripod?


How about

You Tube

You Tube

? Shot by my favorite DoP and my favorite director. Or

You Tube

You Tube

I'd say shoulder mount in most of the Bourne shots.

---

Last edited by kadajawi; 11-30-2013 at 11:03 AM.

[kadajawi]

software fixes like that wreck the resolution, it's terrible.

i agree that optical stabilization helps some things, in fact it's a must-have for some types of shots, but ultimately there isn't any fix for rolling shutter... you have to shoot around it.

Please watch the shots I have uploaded at the link I gave earlier. Please look for rolling shutter. The only cases that I found were when I quickly panned, maybe you can point out more. But the image simply doesn't wobble around. So mechanical IS DOES remove at least most of the wobbling around, simply because what arrives at the sensor isn't moving in a way that creates the wobbling effect that irritates so much (i.e. for the sensor it looks like there is no movement at all). On the other hand electronic IS only INCREASES the wobbling, because all the shakiness arrives on the sensor, the rolling shutter effect is captured (and rolling shutter is worst when the movement is fast and random, left, right, up and down), and then the shakiness is removed which would hide some of the wobbling. That's why people with cameras with electronic SR tend to deactivate SR. Because it WORSENS the rolling shutter effect. And it also has other artifacts, like the steady video getting motion blur for a frame or two (unless you shoot at fast shutter speeds).


I'll give it another try. With electronic SR the sensor captures video with short random bursts of motion blur due to the shaking, and with short random bursts of rolling shutter in various directions, again due to the shaking. And there is no way to undo all of that. But if the video is stabilized in the lens or by moving the sensor, what the sensor captures is not moving at all. It is perfectly stable. And THAT is the big advantage of the K-5, THAT is the reason why we want to be able to use the sensor shift based SR system, instead of an electronic SR system. Or be able to use both, either seperately depending on need, or combined, or none at all. I'd be all for the option of correcting roll using an electronic SR system, while the rest is done by moving the sensor.


The major downside of sensor based stabilization is noise and that it moves back after pans. Noise can be fought with by using an external mic... it's not like the built in mic is good anyway... and even with the built in one I must say it rarely disturbs me. Moving back after pans... shouldn't that happen anyway, even with an electronic system? It's also something Pentax could tweak in software, for example when the camera detects that you are panning horizontally it will mostly stabilize on the vertical axis, leaving it centered on the horizontal axis. That way when you stop panning the sensor doesn't have to move back.

[osv]

that wong kar wai stuff was really good... i think dolly and wide-angle steadicam... the slow crawl stuff was xlnt.

bourne was all steadicam, the horizon line was always stable.

no hand-held that i could see... i told that guy to use a tripod because that's where everyone should start to learn how to shoot video.

this is how it's done professionally, using a dslr, 5d markii:

"I don't use any kind of autofocus while shooting video. In fact, you can't do it. I shot everything in manual and carefully measured and stayed inside the focus area of the lens without having to refocus. It was impossible to manually shift focus on the crane and Glidecam. For the couple fighting at the beginning of the video, I used a Canon 50mm f1.4 just about open and kept my fingers on the focus ring the whole time. I was attached to a larger monitor and judging focus on the fly.

Aside from the 50mm, I mostly used a 24-105mm f4 lens and for some of the shots on the park bench I used a 70-200mm f4.

...We used LED panels to light the face in outdoor daylight shots. For the interiors, we used Arri spots and a large white sheet to soften the light and also bounced some off the ceiling. 1k and two 500w lights I believe. I have no idea what the ISO was."

You Tube

You Tube

[osv]

Please watch the shots I have uploaded at the link I gave earlier. Please look for rolling shutter.

good job on that restaurant shot!! wide angle, shutter speed set to blur the jellocam? stabilization definitely helped there... the only criticism i'd have is that the footage is too soft, even at 1080p... but sometimes that's the look that's called for.

now try to duplicate that killer wong kar wai slow crawl shot while hand-holding... :-0 dang, gotta get a tripod! and a parfocal lens...

thanks for the post.

here is the only thing i found on the borne shakycam, looks like the camera guy is wearing a belt of some sort?

http://www.davidbordwell.net/blog/wp-content/uploads/bourne-1.jpg

[osv]

here is an interesting link on the effects of motion blur and framerate... keep clicking the preset links in the middle of the page, it'll change backgrounds and such for every click

Compare frames per second: which looks better?

[kadajawi]

good job on that restaurant shot!! wide angle, shutter speed set to blur the jellocam? stabilization definitely helped there... the only criticism i'd have is that the footage is too soft, even at 1080p... but sometimes that's the look that's called for.

now try to duplicate that killer wong kar wai slow crawl shot while hand-holding... :-0 dang, gotta get a tripod! and a parfocal lens...

thanks for the post.

here is the only thing i found on the borne shakycam, looks like the camera guy is wearing a belt of some sort?

http://www.davidbordwell.net/blog/wp-content/uploads/bourne-1.jpg

Thanks! The shutter speed was set to whatever the K-5 liked, cause, well, it's a K-5. Can't remember what aperture I used, I might have stopped it down to 4.5 or so to get at least slightly more DoF. But yes, the footage is too soft, and YouTube doesn't help. I doubt the shutter speed blurred the jellocam away, if I get the chance to I'll try something like that in bright daylight to get a short shutter speed. Or maybe I'll find something on my hard drive. Before that shot I used to have problems with the up and down motion while walking, but I didn't notice any jellocam. As I said the K-5 SR does a good job of eliminating it. I'm able of getting some reasonably smooth video walking/shooting with the 50 1.8 too though (the 18-55 at 50mm won't be nearly as smooth...).


The 18-55 kit lens is actually parfocal (at least mine... should be the first 18-55 kit Pentax produced...). However the zoom ring isn't nearly smooth enough to be used during video, and it introduces rotation that you really won't want.


Have you watched the whole WKW clip? After 1:30 it looks handheld to me. The camera is doing some crazy stuff in tight spaces, and isn't really steady. It really all depends on what you want to get.


As for Bourne... maybe I'm wrong, but shouldn't the footage be more stable when shooting with a steadicam rig? The camera moves pretty fast, it bobs up and down when the cameraman is walking, ... not what I'd typically associate with steadicam. But it looks similar to sample footage from a shoulder mount camera I've seen. Also there is one shot where the camera rotates around Bourne. The horizon line doesn't stay stable there.


And the Tyler Stenson video... I think I could get most of the shots without a glidecam (except for the crane shots and what looks like a slider?). Especially the walking and singing parts are no problem with the K-5 SR. Some of the handheld shots look like more of a challenge... the SR would do too good a job of smoothing those out, so I'd probably have to shoot them without SR and just hope there isn't too much microjudder (which often plagues videos shot on Canon DSLRs).


Btw., if you aren't familiar with Wong Kar Wai, it's time to change that. Great filmmaker, especially when he worked with Christopher Doyle. His best works were probably Chungking Express and In the Mood for Love, both having a rather different visual style. Fallen Angels pushes the handheld wide angle aesthetics to the max, while 2046 is just gorgeous, well composed, ...

[kadajawi]

As for pro's not shooting handheld... watch this footage, aired on German television, by a big station, on a relatively widely watched show.

You Tube

You Tube

A K-5 would have done a great job there turning it into something not jellocam-y. And it's not the first time I've seen stuff like that on TV. They like to shoot DSLRs these days it seems, but they shoot Canon, which isn't stabilized in body. Looks great, until there is some movement. And there is. This was a run and gun thing where they probably didn't have permissions (look at all the cops and security forces appearing), shooting in front of a NSA building. In body or optical stabilization would have helped a lot.

[TomGarn]

Now are the Tamron VC lenses are as good or better than the built in SR of a K-3/5? If the working area is even smaller than that of the Pentax I'd tend to answer with no... though you'd have to test it properly first.

Yes, Tamron VC-area subjectively feels a little smaller (for me !) but it is safe, so I like to call
it in fact much more reliable. And there is no modification of my panning needed. But I will watch
out now if I can find any faults in my Tamron-Canon-jobs in the future (never saw one yet, about VC)

And also I will see if I can manage to pan with anti-yello in my muscle controle ;-) with "breaks on"
as long as I use my old K-7. The K-5 videos i saw unto now seem to look better but not really sharp
and crisp as well, but that may be because of other reasons - and I didn't look so very close to it yet.

Btw.: You all just talk about K-5 videos. Is that because you neglect differences between K-5 and K-5II
because there are no differences really - and you just call it K-5 not to make a useless distinction, or
is it because you all mostly got a K-5 ? I wonder about that sometimes.
The noise at least should be less in the K-5II. Isn't it ?

[TomGarn]

Handheld-shot movie ... ;-)

You Tube

You Tube

[kadajawi]

Ah, I must say I don't remember much of A bout de souffle. Except for Belmondo's lips. It's been a while. But what a trailer. They don't make them like they used to...


Mh... it depends, I think. When I shoot at f8 it does look quite sharp, or when I use my 50mm. I guess the kit lens just doesn't do that good a job... I do also have completely deactivated sharpening in the camera. I don't like that sharpened look with halos that other cameras like to produce.


As for K-5... I simply own a K-5, but I don't think there were any differences between the K-5 and the K-5 II in terms of image/video quality. The K-5 IIs was a bit different though... not sure how much that would show in video (aliasing being even worse...?).


I'm uploading another recent video where I am walking at a faster pace, on a sunny day. It is much sharper and shot at a higher shutter speed. This is a shot where I expect the K-3 to fall apart in terms of compression... I'm using crf 20.5, Hi10p (10 bit encode) with the placebo setting, and it resulted in a 25 Mbps file. Considering that the K-3 will have a similar bitrate, but with much less efficient encoding... Anyway, I'll be updating this post once it has uploaded...

You Tube

You Tube

Sadly YouTube has compressed the hell out of the video... it was really quite sharp before YT destroyed it.


I processed the shot with the warp stabilizer, and it looks pretty damn good. No blurrying, no jellocam. I suppose it was already smooth enough (without motion blur due to the shutter speed) so that the warp stabilizer had an easy job fixing it.

[osv]

Sadly YouTube has compressed the hell out of the video... it was really quite sharp before YT destroyed it.

youtube actually does a pretty good job of it, but you have to throw some bitrate at the upload... did you use two-pass encoding? what bitrate was it encoded at?

[kadajawi]

youtube actually does a pretty good job of it, but you have to throw some bitrate at the upload... did you use two-pass encoding? what bitrate was it encoded at?

Single pass crf 20.5, it resulted in a 25 Mbps file that looks as good as the MJPEG. What I uploaded looked good. What YouTube made of it didn't... there's a ton of detail that is constantly changing, so a rather challenging scene.

[osv]

it is a tough scene, but 25Mbps should have been enough... what you really need a good two-pass encoder that does h.264.

the way to test your settings/encoder quality is to put a text or graphic overlay over some of the video... if it's not fairly sharp on the edges after the youtube upload, you know that the encoding job that was done on the upload was not up to par, because the graphic is not relevant to motion in the video.

if anything, since the graphic doesn't change in the frame, it leaves more bits to be allocated to the video footage itself.

i would also compare the youtube upload results both with and without the warp stabilizer, at the same upload bitrate settings for both... my prediction is that the non-warped upload will look cleaner, depending on how bad the footage was to begin with.

[TomGarn]

The combined both SR's look quite good - at times that just looks as smooth as a steadycam,
but then again the flow stops here and there as if some continuity whatever is stuttering .. lol ...
but that maybe the codec, the bitrate, the whatever - low rate, yes ...

I am proud of my Premiere CS6 now - but my knwoledge about importing and exporting the very
best ways is still very shallow and I tend to learn each time just a little bit - but I am a lazy student.

Last time I uploaded that swamp-forest-villa-SR-Outtakes-thing to vimeo for a first time at all -
I up to now always used Tube for my videos - now some info-text popped up saying that this quality
most probably was not good enough because the used bitrate was to small. I checked and had trusted
some stupid automatic set-up exporting the video out of Premiere without a closer look into the details ...
I changed that and put it up again with 150% more bitrate ...

The next day I saw - when watching a video tutorial - that I still had done another minor mistake
while importing the file and not asking for the right sequence-set-up then. Now in Premiere one can
import the file first and then say make a right sequence adapting to this file - and then look if it is right
and even do some more changes ... I still do have my doubts and the proof I sometimes don't make
the best out of my files ... that is really annoying ... and this decoding / encoding stuff urgently needs
better understanding ...

Premiere CS6 now can easily import MJPEG and also XDCAM422 - so I hope I do some good steps
now to manage better: But guess what ? Now I can't import my old 240p 15fps .mov-files anymore
Premiere thinks it's a sound file only when importing them ... So this is the flow of progress, lol - and
I can't work on my first consumer sound-videos anymore I did with my good old Pentax Optio 430.

For to get a good used GH3 with a 12-35 / 2.8 still around 1600,- € have to be payed - so I cling to the
low-budget-hope to either get a good and cheap K-5II or gain the optimism and money for a new K-3.
Those guys with Lumixes do great jobs often enough ... for there is budget and fun ...
https://vimeo.com/62274504
And I feel again ... what it means to be a Pentaxian

[kadajawi]

it is a tough scene, but 25Mbps should have been enough... what you really need a good two-pass encoder that does h.264.

the way to test your settings/encoder quality is to put a text or graphic overlay over some of the video... if it's not fairly sharp on the edges after the youtube upload, you know that the encoding job that was done on the upload was not up to par, because the graphic is not relevant to motion in the video.

if anything, since the graphic doesn't change in the frame, it leaves more bits to be allocated to the video footage itself.

i would also compare the youtube upload results both with and without the warp stabilizer, at the same upload bitrate settings for both... my prediction is that the non-warped upload will look cleaner, depending on how bad the footage was to begin with.

This is part of my archiving script, otherwise I'd need a ton of new hard drives...
avs2yuv "%%~nA.avs" -o - | x264-10b-r2377-1ca7bb9.exe --preset placebo --tune film --qcomp 0.65 --aq-mode=2 --crf 20.5 --fps 25 --input-res 1920x1080 --output "%%~nA.264" - --demuxer y4m


I'm using the latest 10 bit version of x264, at the very highest setting. Constant quality rather than VBR, cause some scenes can easily get away at 2 Mbps, while others need 25 (usually I get 6-8, though that also depends on the lens. The 50mm usually gets smaller files). Only setting the bitrate doesn't make sense to me. I do a color denoise prior to encoding on all files (pretty much like what Lightroom does per default), and a mild denoise in general. If you have any suggestion on where to upload the file I uploaded to YouTube I will do so.


I have attached 2 screen grabs, one from the original and one from the h264 file. I think the 2 star 60 Mbps MJPEG setting wasn't enough for a scene like that, sadly I didn't bring enough memory along to shoot at the highest quality setting all the time. In situations like that it'd be nice to have h264... which the K-3 does, but sadly it's the only format it supports. Rather than MJPEG they could let us shoot h264 at really high bitrates. Say 50-80 Mbps. Or have the encoder be more open ended, so if the scene needs it it can give much higher bitrates. The Sony a57 simply warns you that you shoot at a bitrate/setting that isn't Bluray compatible... why can't Pentax do something like that? Simply do a here be dragons pop up when you set something that CAN have downsides, such as very high bitrates. Say it needs to be processed, cause the bitrate is too high for any standard medium. Say, when selecting mechanical SR, that it can introduce noise in the audio track, and that it is advisable to shoot with a microphone that is isolated from the camera. And that it reduces battery life.


The warped footage looks pretty clean to me. There is a slight loss in resolution, but since there wasn't much to correct for - just a bit of jitter mostly - it didn't crop much. I think something like 10%?


@TomGarn: You can always convert the old footage to something more acceptable for Premiere, and since the resolution is so low you can go for something lossless like Huffyuv or better yet Lagarith.


As for the flow stopping here and there... either it's the low frame rate that you are talking about, or I have simply maxed out the range of the SR system when walking fast (I can't get stable video while running, for that you really need at least a glidecam). My dream would have been for the K-3 to cover the entire full frame area, and of course for it to do video SR. With such a range (and more rotational adjustments) you would be able to get ridiculously stable video... and to bracket a full frame photo at high resolution (i.e. the sensor would move to the top left, shoot, top right, shoot, bottom left, shoot, bottom right, shoot). That could be loaded into any stitching software and voila you have a very high res full frame image. Landscapers and architectural photographers rejoice! Alas that never happened...


I also looked at the Vimeo preset in Premiere, and saw 5 Mbps (!). The YouTube preset gives you 8 Mbps... still too low, but since YouTube will process the video anyway, and won't do such a good job at it (not surprisingly, encoding takes a lot of CPU power at high settings, and decoding those files is also quite challenging, so YouTube just does a quick and dirty encode job... they just have too many files to encode).


The GH3 is really popular for film making because it a) has ridiculously high bitrates when hacked and b) cause it does pixel binning... it takes ALL the pixels and scales them down to Full HD resolution. No aliasing, much higher light sensitivity, sharp video... it's I think the only camera that does that, perhaps other sensors aren't capable of reading out all pixels 25, 30, 50 times a second.


However there are people doing great work with Pentaxes... like this guy who is also on PF:

Vimeo

I have also seen Film Riot using a Pentax K-5 or K-7 several times, though mostly they will shoot a 7D or 5D it seems, and occasionally a RED. On the review of several glidecam-a-likes they've done a few months back you can see them use the Pentax. I wonder if they left the SR on for that extra bit of stability (just in case some bumps do get through).